1.Pedestrian 3D Bounding Box Prediction ⬇️
Safety is still the main issue of autonomous driving, and in order to be globally deployed, they need to predict pedestrians' motions sufficiently in advance. While there is a lot of research on coarse-grained (human center prediction) and fine-grained predictions (human body keypoints prediction), we focus on 3D bounding boxes, which are reasonable estimates of humans without modeling complex motion details for autonomous vehicles. This gives the flexibility to predict in longer horizons in real-world settings. We suggest this new problem and present a simple yet effective model for pedestrians' 3D bounding box prediction. This method follows an encoder-decoder architecture based on recurrent neural networks, and our experiments show its effectiveness in both the synthetic (JTA) and real-world (NuScenes) datasets. The learned representation has useful information to enhance the performance of other tasks, such as action anticipation. Our code is available online: this https URL
2.High-Resolution Virtual Try-On with Misalignment and Occlusion-Handled Conditions ⬇️
Image-based virtual try-on aims to synthesize an image of a person wearing a given clothing item. To solve the task, the existing methods warp the clothing item to fit the person's body and generate the segmentation map of the person wearing the item, before fusing the item with the person. However, when the warping and the segmentation generation stages operate individually without information exchange, the misalignment between the warped clothes and the segmentation map occurs, which leads to the artifacts in the final image. The information disconnection also causes excessive warping near the clothing regions occluded by the body parts, so called pixel-squeezing artifacts. To settle the issues, we propose a novel try-on condition generator as a unified module of the two stages (i.e., warping and segmentation generation stages). A newly proposed feature fusion block in the condition generator implements the information exchange, and the condition generator does not create any misalignment or pixel-squeezing artifacts. We also introduce discriminator rejection that filters out the incorrect segmentation map predictions and assures the performance of virtual try-on frameworks. Experiments on a high-resolution dataset demonstrate that our model successfully handles the misalignment and the occlusion, and significantly outperforms the baselines. Code is available at this https URL.
3.Visualizing and Alleviating the Effect of Radial Distortion on Camera Calibration Using Principal Lines ⬇️
Preparing appropriate images for camera calibration is crucial to obtain accurate results. In this paper, new suggestions for preparing such data to alleviate the adverse effect of radial distortion for a calibration procedure using principal lines are developed through the investigations of: (i) identifying directions of checkerboard movements in an image which will result in maximum (and minimum) influence on the calibration results, and (ii) inspecting symmetry and monotonicity of such effect in (i) using the above principal lines. Accordingly, it is suggested that the estimation of principal point should based on linearly independent pairs of nearly parallel principal lines, with a member in each pair corresponds to a near 180-degree rotation (in the image plane) of the other. Experimental results show that more robust and consistent calibration results for the foregoing estimation can actually be obtained, compared with the renowned algebraic methods which estimate distortion parameters explicitly.
4.SSL-Lanes: Self-Supervised Learning for Motion Forecasting in Autonomous Driving ⬇️
Self-supervised learning (SSL) is an emerging technique that has been successfully employed to train convolutional neural networks (CNNs) and graph neural networks (GNNs) for more transferable, generalizable, and robust representation learning. However its potential in motion forecasting for autonomous driving has rarely been explored. In this study, we report the first systematic exploration and assessment of incorporating self-supervision into motion forecasting. We first propose to investigate four novel self-supervised learning tasks for motion forecasting with theoretical rationale and quantitative and qualitative comparisons on the challenging large-scale Argoverse dataset. Secondly, we point out that our auxiliary SSL-based learning setup not only outperforms forecasting methods which use transformers, complicated fusion mechanisms and sophisticated online dense goal candidate optimization algorithms in terms of performance accuracy, but also has low inference time and architectural complexity. Lastly, we conduct several experiments to understand why SSL improves motion forecasting. Code is open-sourced at \url{this https URL}.
5.Rethinking Adversarial Examples for Location Privacy Protection ⬇️
We have investigated a new application of adversarial examples, namely location privacy protection against landmark recognition systems. We introduce mask-guided multimodal projected gradient descent (MM-PGD), in which adversarial examples are trained on different deep models. Image contents are protected by analyzing the properties of regions to identify the ones most suitable for blending in adversarial examples. We investigated two region identification strategies: class activation map-based MM-PGD, in which the internal behaviors of trained deep models are targeted; and human-vision-based MM-PGD, in which regions that attract less human attention are targeted. Experiments on the Places365 dataset demonstrated that these strategies are potentially effective in defending against black-box landmark recognition systems without the need for much image manipulation.
6.Taxonomy and evolution predicting using deep learning in images ⬇️
Molecular and morphological characters, as important parts of biological taxonomy, are contradictory but need to be integrated. Organism's image recognition and bioinformatics are emerging and hot problems nowadays but with a gap between them. In this work, a multi-branching recognition framework mediated by genetic information bridges this barrier, which establishes the link between macro-morphology and micro-molecular information of mushrooms. The novel multi-perspective structure is proposed to fuse the feature images from three branching models, which significantly improves the accuracy of recognition by about 10% and up to more than 90%. Further, genetic information is implemented to the mushroom image recognition task by using genetic distance embeddings as the representation space for predicting image distance and species identification. Semantic overfitting of traditional classification tasks and the granularity of fine-grained image recognition are also discussed in depth for the first time. The generalizability of the model was investigated in fine-grained scenarios using zero-shot learning tasks, which could predict the taxonomic and evolutionary information of unseen samples. We presented the first method to map images to DNA, namely used an encoder mapping image to genetic distances, and then decoded DNA through a pre-trained decoder, where the total test accuracy on 37 species for DNA prediction is 87.45%. This study creates a novel recognition framework by systematically studying the mushroom image recognition problem, bridging the gap between macroscopic biological information and microscopic molecular information, which will provide a new reference for intelligent biometrics in the future.
7.Show Me Your Face, And I'll Tell You How You Speak ⬇️
When we speak, the prosody and content of the speech can be inferred from the movement of our lips. In this work, we explore the task of lip to speech synthesis, i.e., learning to generate speech given only the lip movements of a speaker where we focus on learning accurate lip to speech mappings for multiple speakers in unconstrained, large vocabulary settings. We capture the speaker's voice identity through their facial characteristics, i.e., age, gender, ethnicity and condition them along with the lip movements to generate speaker identity aware speech. To this end, we present a novel method "Lip2Speech", with key design choices to achieve accurate lip to speech synthesis in unconstrained scenarios. We also perform various experiments and extensive evaluation using quantitative, qualitative metrics and human evaluation.
8.Detecting tiny objects in aerial images: A normalized Wasserstein distance and a new benchmark ⬇️
Tiny object detection (TOD) in aerial images is challenging since a tiny object only contains a few pixels. State-of-the-art object detectors do not provide satisfactory results on tiny objects due to the lack of supervision from discriminative features. Our key observation is that the Intersection over Union (IoU) metric and its extensions are very sensitive to the location deviation of the tiny objects, which drastically deteriorates the quality of label assignment when used in anchor-based detectors. To tackle this problem, we propose a new evaluation metric dubbed Normalized Wasserstein Distance (NWD) and a new RanKing-based Assigning (RKA) strategy for tiny object detection. The proposed NWD-RKA strategy can be easily embedded into all kinds of anchor-based detectors to replace the standard IoU threshold-based one, significantly improving label assignment and providing sufficient supervision information for network training. Tested on four datasets, NWD-RKA can consistently improve tiny object detection performance by a large margin. Besides, observing prominent noisy labels in the Tiny Object Detection in Aerial Images (AI-TOD) dataset, we are motivated to meticulously relabel it and release AI-TOD-v2 and its corresponding benchmark. In AI-TOD-v2, the missing annotation and location error problems are considerably mitigated, facilitating more reliable training and validation processes. Embedding NWD-RKA into DetectoRS, the detection performance achieves 4.3 AP points improvement over state-of-the-art competitors on AI-TOD-v2. Datasets, codes, and more visualizations are available at: this https URL
9.Learning Gait Representation from Massive Unlabelled Walking Videos: A Benchmark ⬇️
Gait depicts individuals' unique and distinguishing walking patterns and has become one of the most promising biometric features for human identification. As a fine-grained recognition task, gait recognition is easily affected by many factors and usually requires a large amount of completely annotated data that is costly and insatiable. This paper proposes a large-scale self-supervised benchmark for gait recognition with contrastive learning, aiming to learn the general gait representation from massive unlabelled walking videos for practical applications via offering informative walking priors and diverse real-world variations. Specifically, we collect a large-scale unlabelled gait dataset GaitLU-1M consisting of 1.02M walking sequences and propose a conceptually simple yet empirically powerful baseline model GaitSSB. Experimentally, we evaluate the pre-trained model on four widely-used gait benchmarks, CASIA-B, OU-MVLP, GREW and Gait3D with or without transfer learning. The unsupervised results are comparable to or even better than the early model-based and GEI-based methods. After transfer learning, our method outperforms existing methods by a large margin in most cases. Theoretically, we discuss the critical issues for gait-specific contrastive framework and present some insights for further study. As far as we know, GaitLU-1M is the first large-scale unlabelled gait dataset, and GaitSSB is the first method that achieves remarkable unsupervised results on the aforementioned benchmarks. The source code of GaitSSB will be integrated into OpenGait which is available at this https URL.
10.Primitive Graph Learning for Unified Vector Mapping ⬇️
Large-scale vector mapping is important for transportation, city planning, and survey and census. We propose GraphMapper, a unified framework for end-to-end vector map extraction from satellite images. Our key idea is a novel unified representation of shapes of different topologies named "primitive graph", which is a set of shape primitives and their pairwise relationship matrix. Then, we convert vector shape prediction, regularization, and topology reconstruction into a unique primitive graph learning problem. Specifically, GraphMapper is a generic primitive graph learning network based on global shape context modelling through multi-head-attention. An embedding space sorting method is developed for accurate primitive relationship modelling. We empirically demonstrate the effectiveness of GraphMapper on two challenging mapping tasks, building footprint regularization and road network topology reconstruction. Our model outperforms state-of-the-art methods by 8-10% in both tasks on public benchmarks. All code will be publicly available.
11.Multi-Prior Learning via Neural Architecture Search for Blind Face Restoration ⬇️
Blind Face Restoration (BFR) aims to recover high-quality face images from low-quality ones and usually resorts to facial priors for improving restoration performance. However, current methods still suffer from two major difficulties: 1) how to derive a powerful network architecture without extensive hand tuning; 2) how to capture complementary information from multiple facial priors in one network to improve restoration performance. To this end, we propose a Face Restoration Searching Network (FRSNet) to adaptively search the suitable feature extraction architecture within our specified search space, which can directly contribute to the restoration quality. On the basis of FRSNet, we further design our Multiple Facial Prior Searching Network (MFPSNet) with a multi-prior learning scheme. MFPSNet optimally extracts information from diverse facial priors and fuses the information into image features, ensuring that both external guidance and internal features are reserved. In this way, MFPSNet takes full advantage of semantic-level (parsing maps), geometric-level (facial heatmaps), reference-level (facial dictionaries) and pixel-level (degraded images) information and thus generates faithful and realistic images. Quantitative and qualitative experiments show that MFPSNet performs favorably on both synthetic and real-world datasets against the state-of-the-art BFR methods. The codes are publicly available at: this https URL.
12.Robustifying Vision Transformer without Retraining from Scratch by Test-Time Class-Conditional Feature Alignment ⬇️
Vision Transformer (ViT) is becoming more popular in image processing. Specifically, we investigate the effectiveness of test-time adaptation (TTA) on ViT, a technique that has emerged to correct its prediction during test-time by itself. First, we benchmark various test-time adaptation approaches on ViT-B16 and ViT-L16. It is shown that the TTA is effective on ViT and the prior-convention (sensibly selecting modulation parameters) is not necessary when using proper loss function. Based on the observation, we propose a new test-time adaptation method called class-conditional feature alignment (CFA), which minimizes both the class-conditional distribution differences and the whole distribution differences of the hidden representation between the source and target in an online manner. Experiments of image classification tasks on common corruption (CIFAR-10-C, CIFAR-100-C, and ImageNet-C) and domain adaptation (digits datasets and ImageNet-Sketch) show that CFA stably outperforms the existing baselines on various datasets. We also verify that CFA is model agnostic by experimenting on ResNet, MLP-Mixer, and several ViT variants (ViT-AugReg, DeiT, and BeiT). Using BeiT backbone, CFA achieves 19.8% top-1 error rate on ImageNet-C, outperforming the existing test-time adaptation baseline 44.0%. This is a state-of-the-art result among TTA methods that do not need to alter training phase.
13.Generating near-infrared facial expression datasets with dimensional affect labels ⬇️
Facial expression analysis has long been an active research area of computer vision. Traditional methods mainly analyse images for prototypical discrete emotions; as a result, they do not provide an accurate depiction of the complex emotional states in humans. Furthermore, illumination variance remains a challenge for face analysis in the visible light spectrum. To address these issues, we propose using a dimensional model based on valence and arousal to represent a wider range of emotions, in combination with near infra-red (NIR) imagery, which is more robust to illumination changes. Since there are no existing NIR facial expression datasets with valence-arousal labels available, we present two complementary data augmentation methods (face morphing and CycleGAN approach) to create NIR image datasets with dimensional emotion labels from existing categorical and/or visible-light datasets. Our experiments show that these generated NIR datasets are comparable to existing datasets in terms of data quality and baseline prediction performance.
14.Accurate and Real-time Pseudo Lidar Detection: Is Stereo Neural Network Really Necessary? ⬇️
The proposal of Pseudo-Lidar representation has significantly narrowed the gap between visual-based and active Lidar-based 3D object detection. However, current researches exclusively focus on pushing the accuracy improvement of Pseudo-Lidar by taking the advantage of complex and time-consuming neural networks. Seldom explore the profound characteristics of Pseudo-Lidar representation to obtain the promoting opportunities. In this paper, we dive deep into the pseudo Lidar representation and argue that the performance of 3D object detection is not fully dependent on the high precision stereo depth estimation. We demonstrate that even for the unreliable depth estimation, with proper data processing and refining, it can achieve comparable 3D object detection accuracy. With this finding, we further show the possibility that utilizing fast but inaccurate stereo matching algorithms in the Pseudo-Lidar system to achieve low latency responsiveness. In the experiments, we develop a system with a less powerful stereo matching predictor and adopt the proposed refinement schemes to improve the accuracy. The evaluation on the KITTI benchmark shows that the presented system achieves competitive accuracy to the state-of-the-art approaches with only 23 ms computing, showing it is a suitable candidate for deploying to real car-hold applications.
15.When the Sun Goes Down: Repairing Photometric Losses for All-Day Depth Estimation ⬇️
Self-supervised deep learning methods for joint depth and ego-motion estimation can yield accurate trajectories without needing ground-truth training data. However, as they typically use photometric losses, their performance can degrade significantly when the assumptions these losses make (e.g. temporal illumination consistency, a static scene, and the absence of noise and occlusions) are violated. This limits their use for e.g. nighttime sequences, which tend to contain many point light sources (including on dynamic objects) and low signal-to-noise ratio (SNR) in darker image regions. In this paper, we show how to use a combination of three techniques to allow the existing photometric losses to work for both day and nighttime images. First, we introduce a per-pixel neural intensity transformation to compensate for the light changes that occur between successive frames. Second, we predict a per-pixel residual flow map that we use to correct the reprojection correspondences induced by the estimated ego-motion and depth from the networks. And third, we denoise the training images to improve the robustness and accuracy of our approach. These changes allow us to train a single model for both day and nighttime images without needing separate encoders or extra feature networks like existing methods. We perform extensive experiments and ablation studies on the challenging Oxford RobotCar dataset to demonstrate the efficacy of our approach for both day and nighttime sequences.
16.Cross-Forgery Analysis of Vision Transformers and CNNs for Deepfake Image Detection ⬇️
Deepfake Generation Techniques are evolving at a rapid pace, making it possible to create realistic manipulated images and videos and endangering the serenity of modern society. The continual emergence of new and varied techniques brings with it a further problem to be faced, namely the ability of deepfake detection models to update themselves promptly in order to be able to identify manipulations carried out using even the most recent methods. This is an extremely complex problem to solve, as training a model requires large amounts of data, which are difficult to obtain if the deepfake generation method is too recent. Moreover, continuously retraining a network would be unfeasible. In this paper, we ask ourselves if, among the various deep learning techniques, there is one that is able to generalise the concept of deepfake to such an extent that it does not remain tied to one or more specific deepfake generation methods used in the training set. We compared a Vision Transformer with an EfficientNetV2 on a cross-forgery context based on the ForgeryNet dataset. From our experiments, It emerges that EfficientNetV2 has a greater tendency to specialize often obtaining better results on training methods while Vision Transformers exhibit a superior generalization ability that makes them more competent even on images generated with new methodologies.
17.FedRare: Federated Learning with Intra- and Inter-Client Contrast for Effective Rare Disease Classification ⬇️
Federated learning (FL), enabling different medical institutions or clients to train a model collaboratively without data privacy leakage, has drawn great attention in medical imaging communities recently. Though inter-client data heterogeneity has been thoroughly studied, the class imbalance problem due to the existence of rare diseases still is under-explored. In this paper, we propose a novel FL framework FedRare for medical image classification especially on dealing with data heterogeneity with the existence of rare diseases. In FedRare, each client trains a model locally to extract highly-separable latent features for classification via intra-client supervised contrastive learning. Considering the limited data on rare diseases, we build positive sample queues for augmentation (i.e. data re-sampling). The server in FedRare would collect the latent features from clients and automatically select the most reliable latent features as guidance sent back to clients. Then, each client is jointly trained by an inter-client contrastive loss to align its latent features to the federated latent features of full classes. In this way, the parameter/feature variances across clients are effectively minimized, leading to better convergence and performance improvements. Experimental results on the publicly-available dataset for skin lesion diagnosis demonstrate FedRare's superior performance. Under the 10-client federated setting where four clients have no rare disease samples, FedRare achieves an average increase of 9.60% and 5.90% in balanced accuracy compared to the baseline framework FedAvg and the state-of-the-art approach FedIRM respectively. Considering the board existence of rare diseases in clinical scenarios, we believe FedRare would benefit future FL framework design for medical image classification. The source code of this paper is publicly available at this https URL.
18.3D Multi-Object Tracking with Differentiable Pose Estimation ⬇️
We propose a novel approach for joint 3D multi-object tracking and reconstruction from RGB-D sequences in indoor environments. To this end, we detect and reconstruct objects in each frame while predicting dense correspondences mappings into a normalized object space. We leverage those correspondences to inform a graph neural network to solve for the optimal, temporally-consistent 7-DoF pose trajectories of all objects. The novelty of our method is two-fold: first, we propose a new graph-based approach for differentiable pose estimation over time to learn optimal pose trajectories; second, we present a joint formulation of reconstruction and pose estimation along the time axis for robust and geometrically consistent multi-object tracking. In order to validate our approach, we introduce a new synthetic dataset comprising 2381 unique indoor sequences with a total of 60k rendered RGB-D images for multi-object tracking with moving objects and camera positions derived from the synthetic 3D-FRONT dataset. We demonstrate that our method improves the accumulated MOTA score for all test sequences by 24.8% over existing state-of-the-art methods. In several ablations on synthetic and real-world sequences, we show that our graph-based, fully end-to-end-learnable approach yields a significant boost in tracking performance.
19.Adversarial Consistency for Single Domain Generalization in Medical Image Segmentation ⬇️
An organ segmentation method that can generalize to unseen contrasts and scanner settings can significantly reduce the need for retraining of deep learning models. Domain Generalization (DG) aims to achieve this goal. However, most DG methods for segmentation require training data from multiple domains during training. We propose a novel adversarial domain generalization method for organ segmentation trained on data from a \emph{single} domain. We synthesize the new domains via learning an adversarial domain synthesizer (ADS) and presume that the synthetic domains cover a large enough area of plausible distributions so that unseen domains can be interpolated from synthetic domains. We propose a mutual information regularizer to enforce the semantic consistency between images from the synthetic domains, which can be estimated by patch-level contrastive learning. We evaluate our method for various organ segmentation for unseen modalities, scanning protocols, and scanner sites.
20.A Comprehensive Survey on Deep Gait Recognition: Algorithms, Datasets and Challenges ⬇️
Gait recognition aims at identifying a person at a distance through visual cameras. With the emergence of deep learning, significant advancements in gait recognition have achieved inspiring success in many scenarios by utilizing deep learning techniques. Nevertheless, the increasing need for video surveillance introduces more challenges, including robust recognition under various variances, modeling motion information in gait sequences, unfair performance comparison due to protocol variances, biometrics security, and privacy prevention. This paper provides a comprehensive survey of deep learning for gait recognition. We first present the odyssey of gait recognition from traditional algorithms to deep models, providing explicit knowledge of the whole workflow of a gait recognition system. Then deep learning for gait recognition is discussed from the perspective of deep representations and architecture with an in-depth summary. Specifically, deep gait representations are categorized into static and dynamic features, while deep architectures include single-stream and multi-stream architecture. Following our proposed taxonomy with novelty, it can be beneficial for providing inspiration and promoting the perception of deep gait recognition. Besides, we also present a comprehensive summary of all vision-based gait datasets and the performance analysis. Finally, the article discusses some open issues with significant potential prospects.
21.Boosting R-CNN: Reweighting R-CNN Samples by RPN's Error for Underwater Object Detection ⬇️
Complicated underwater environments bring new challenges to object detection, such as unbalanced light conditions, low contrast, occlusion, and mimicry of aquatic organisms. Under these circumstances, the objects captured by the underwater camera will become vague, and the generic detectors often fail on these vague objects. This work aims to solve the problem from two perspectives: uncertainty modeling and hard example mining. We propose a two-stage underwater detector named boosting R-CNN, which comprises three key components. First, a new region proposal network named RetinaRPN is proposed, which provides high-quality proposals and considers objectness and IoU prediction for uncertainty to model the object prior probability. Second, the probabilistic inference pipeline is introduced to combine the first-stage prior uncertainty and the second-stage classification score to model the final detection score. Finally, we propose a new hard example mining method named boosting reweighting. Specifically, when the region proposal network miscalculates the object prior probability for a sample, boosting reweighting will increase the classification loss of the sample in the R-CNN head during training, while reducing the loss of easy samples with accurately estimated priors. Thus, a robust detection head in the second stage can be obtained. During the inference stage, the R-CNN has the capability to rectify the error of the first stage to improve the performance. Comprehensive experiments on two underwater datasets and two generic object detection datasets demonstrate the effectiveness and robustness of our method.
22.The Third Place Solution for CVPR2022 AVA Accessibility Vision and Autonomy Challenge ⬇️
The goal of AVA challenge is to provide vision-based benchmarks and methods relevant to accessibility. In this paper, we introduce the technical details of our submission to the CVPR2022 AVA Challenge. Firstly, we conducted some experiments to help employ proper model and data augmentation strategy for this task. Secondly, an effective training strategy was applied to improve the performance. Thirdly, we integrated the results from two different segmentation frameworks to improve the performance further. Experimental results demonstrate that our approach can achieve a competitive result on the AVA test set. Finally, our approach achieves 63.008%AP@0.50:0.95 on the test set of CVPR2022 AVA Challenge.
23.Towards Global-Scale Crowd+AI Techniques to Map and Assess Sidewalks for People with Disabilities ⬇️
There is a lack of data on the location, condition, and accessibility of sidewalks across the world, which not only impacts where and how people travel but also fundamentally limits interactive mapping tools and urban analytics. In this paper, we describe initial work in semi-automatically building a sidewalk network topology from satellite imagery using hierarchical multi-scale attention models, inferring surface materials from street-level images using active learning-based semantic segmentation, and assessing sidewalk condition and accessibility features using Crowd+AI. We close with a call to create a database of labeled satellite and streetscape scenes for sidewalks and sidewalk accessibility issues along with standardized benchmarks.
24.How Many Events do You Need? Event-based Visual Place Recognition Using Sparse But Varying Pixels ⬇️
Event cameras continue to attract interest due to desirable characteristics such as high dynamic range, low latency, virtually no motion blur, and high energy efficiency. One of the potential applications of event camera research lies in visual place recognition for robot localization, where a query observation has to be matched to the corresponding reference place in the database. In this letter, we explore the distinctiveness of event streams from a small subset of pixels (in the tens or hundreds). We demonstrate that the absolute difference in the number of events at those pixel locations accumulated into event frames can be sufficient for the place recognition task, when pixels that display large variations in the reference set are used. Using such sparse (over image coordinates) but varying (variance over the number of events per pixel location) pixels enables frequent and computationally cheap updates of the location estimates. Furthermore, when event frames contain a constant number of events, our method takes full advantage of the event-driven nature of the sensory stream and displays promising robustness to changes in velocity. We evaluate our proposed approach on the Brisbane-Event-VPR dataset in an outdoor driving scenario, as well as the newly contributed indoor QCR-Event-VPR dataset that was captured with a DAVIS346 camera mounted on a mobile robotic platform. Our results show that our approach achieves competitive performance when compared to several baseline methods on those datasets, and is particularly well suited for compute- and energy-constrained platforms such as interplanetary rovers.
25.Feature Refinement to Improve High Resolution Image Inpainting ⬇️
In this paper, we address the problem of degradation in inpainting quality of neural networks operating at high resolutions. Inpainting networks are often unable to generate globally coherent structures at resolutions higher than their training set. This is partially attributed to the receptive field remaining static, despite an increase in image resolution. Although downscaling the image prior to inpainting produces coherent structure, it inherently lacks detail present at higher resolutions. To get the best of both worlds, we optimize the intermediate featuremaps of a network by minimizing a multiscale consistency loss at inference. This runtime optimization improves the inpainting results and establishes a new state-of-the-art for high resolution inpainting. Code is available at: this https URL.
26.Multi-scale Network with Attentional Multi-resolution Fusion for Point Cloud Semantic Segmentation ⬇️
In this paper, we present a comprehensive point cloud semantic segmentation network that aggregates both local and global multi-scale information. First, we propose an Angle Correlation Point Convolution (ACPConv) module to effectively learn the local shapes of points. Second, based upon ACPConv, we introduce a local multi-scale split (MSS) block that hierarchically connects features within one single block and gradually enlarges the receptive field which is beneficial for exploiting the local context. Third, inspired by HRNet which has excellent performance on 2D image vision tasks, we build an HRNet customized for point cloud to learn global multi-scale context. Lastly, we introduce a point-wise attention fusion approach that fuses multi-resolution predictions and further improves point cloud semantic segmentation performance. Our experimental results and ablations on several benchmark datasets show that our proposed method is effective and able to achieve state-of-the-art performances compared to existing methods.
27.Patch Selection for Melanoma Classification ⬇️
In medical image processing, the most important information is often located on small parts of the image. Patch-based approaches aim at using only the most relevant parts of the image. Finding ways to automatically select the patches is a challenge. In this paper, we investigate two criteria to choose patches: entropy and a spectral similarity criterion. We perform experiments at different levels of patch size. We train a Convolutional Neural Network on the subsets of patches and analyze the training time. We find that, in addition to requiring less preprocessing time, the classifiers trained on the datasets of patches selected based on entropy converge faster than on those selected based on the spectral similarity criterion and, furthermore, lead to higher accuracy. Moreover, patches of high entropy lead to faster convergence and better accuracy than patches of low entropy.
28.Reducing Annotation Need in Self-Explanatory Models for Lung Nodule Diagnosis ⬇️
Feature-based self-explanatory methods explain their classification in terms of human-understandable features. In the medical imaging community, this semantic matching of clinical knowledge adds significantly to the trustworthiness of the AI. However, the cost of additional annotation of features remains a pressing issue. We address this problem by proposing cRedAnno, a data-/annotation-efficient self-explanatory approach for lung nodule diagnosis. cRedAnno considerably reduces the annotation need by introducing self-supervised contrastive learning to alleviate the burden of learning most parameters from annotation, replacing end-to-end training with two-stage training. When training with hundreds of nodule samples and only 1% of their annotations, cRedAnno achieves competitive accuracy in predicting malignancy, meanwhile significantly surpassing most previous works in predicting nodule attributes. Visualisation of the learned space further indicates that the correlation between the clustering of malignancy and nodule attributes coincides with clinical knowledge. Our complete code is open-source available: this https URL.
29.NeuRIS: Neural Reconstruction of Indoor Scenes Using Normal Priors ⬇️
Reconstructing 3D indoor scenes from 2D images is an important task in many computer vision and graphics applications. A main challenge in this task is that large texture-less areas in typical indoor scenes make existing methods struggle to produce satisfactory reconstruction results. We propose a new method, named NeuRIS, for high quality reconstruction of indoor scenes. The key idea of NeuRIS is to integrate estimated normal of indoor scenes as a prior in a neural rendering framework for reconstructing large texture-less shapes and, importantly, to do this in an adaptive manner to also enable the reconstruction of irregular shapes with fine details. Specifically, we evaluate the faithfulness of the normal priors on-the-fly by checking the multi-view consistency of reconstruction during the optimization process. Only the normal priors accepted as faithful will be utilized for 3D reconstruction, which typically happens in the regions of smooth shapes possibly with weak texture. However, for those regions with small objects or thin structures, for which the normal priors are usually unreliable, we will only rely on visual features of the input images, since such regions typically contain relatively rich visual features (e.g., shade changes and boundary contours). Extensive experiments show that NeuRIS significantly outperforms the state-of-the-art methods in terms of reconstruction quality.
30.A View Independent Classification Framework for Yoga Postures ⬇️
Yoga is a globally acclaimed and widely recommended practice for a healthy living. Maintaining correct posture while performing a Yogasana is of utmost importance. In this work, we employ transfer learning from Human Pose Estimation models for extracting 136 key-points spread all over the body to train a Random Forest classifier which is used for estimation of the Yogasanas. The results are evaluated on an in-house collected extensive yoga video database of 51 subjects recorded from 4 different camera angles. We propose a 3 step scheme for evaluating the generalizability of a Yoga classifier by testing it on 1) unseen frames, 2) unseen subjects, and 3) unseen camera angles. We argue that for most of the applications, validation accuracies on unseen subjects and unseen camera angles would be most important. We empirically analyze over three public datasets, the advantage of transfer learning and the possibilities of target leakage. We further demonstrate that the classification accuracies critically depend on the cross validation method employed and can often be misleading. To promote further research, we have made key-points dataset and code publicly available.
31.Parameter-Efficient Image-to-Video Transfer Learning ⬇️
Capitalizing on large pre-trained models for various downstream tasks of interest have recently emerged with promising performance. Due to the ever-growing model size, the standard full fine-tuning based task adaptation strategy becomes prohibitively costly in terms of model training and storage. This has led to a new research direction in parameter-efficient transfer learning. However, existing attempts typically focus on downstream tasks from the same modality (e.g., image understanding) of the pre-trained model. This creates a limit because in some specific modalities, (e.g., video understanding) such a strong pre-trained model with sufficient knowledge is less or not available. In this work, we investigate such a novel cross-modality transfer learning setting, namely parameter-efficient image-to-video transfer learning. To solve this problem, we propose a new Spatio-Temporal Adapter (ST-Adapter) for parameter-efficient fine-tuning per video task. With a built-in spatio-temporal reasoning capability in a compact design, ST-Adapter enables a pre-trained image model without temporal knowledge to reason about dynamic video content at a small (~8%) per-task parameter cost, requiring approximately 20 times fewer updated parameters compared to previous work. Extensive experiments on video action recognition tasks show that our ST-Adapter can match or even outperform the strong full fine-tuning strategy and state-of-the-art video models, whilst enjoying the advantage of parameter efficiency.
32.aSTDP: A More Biologically Plausible Learning ⬇️
Spike-timing dependent plasticity in biological neural networks has been proven to be important during biological learning process. On the other hand, artificial neural networks use a different way to learn, such as Back-Propagation or Contrastive Hebbian Learning. In this work we introduce approximate STDP, a new neural networks learning framework more similar to the biological learning process. It uses only STDP rules for supervised and unsupervised learning, every neuron distributed learn patterns and don' t need a global loss or other supervised information. We also use a numerical way to approximate the derivatives of each neuron in order to better use SDTP learning and use the derivatives to set a target for neurons to accelerate training and testing process. The framework can make predictions or generate patterns in one model without additional configuration. Finally, we verified our framework on MNIST dataset for classification and generation tasks.
33.RevBiFPN: The Fully Reversible Bidirectional Feature Pyramid Network ⬇️
This work introduces the RevSilo, the first reversible module for bidirectional multi-scale feature fusion. Like other reversible methods, RevSilo eliminates the need to store hidden activations by recomputing them. Existing reversible methods, however, do not apply to multi-scale feature fusion and are therefore not applicable to a large class of networks. Bidirectional multi-scale feature fusion promotes local and global coherence and has become a de facto design principle for networks targeting spatially sensitive tasks e.g. HRNet and EfficientDet. When paired with high-resolution inputs, these networks achieve state-of-the-art results across various computer vision tasks, but training them requires substantial accelerator memory for saving large, multi-resolution activations. These memory requirements cap network size and limit progress. Using reversible recomputation, the RevSilo alleviates memory issues while still operating across resolution scales. Stacking RevSilos, we create RevBiFPN, a fully reversible bidirectional feature pyramid network. For classification, RevBiFPN is competitive with networks such as EfficientNet while using up to 19.8x lesser training memory. When fine-tuned on COCO, RevBiFPN provides up to a 2.5% boost in AP over HRNet using fewer MACs and a 2.4x reduction in training-time memory.
34.Continual Learning with Transformers for Image Classification ⬇️
In many real-world scenarios, data to train machine learning models become available over time. However, neural network models struggle to continually learn new concepts without forgetting what has been learnt in the past. This phenomenon is known as catastrophic forgetting and it is often difficult to prevent due to practical constraints, such as the amount of data that can be stored or the limited computation sources that can be used. Moreover, training large neural networks, such as Transformers, from scratch is very costly and requires a vast amount of training data, which might not be available in the application domain of interest. A recent trend indicates that dynamic architectures based on an expansion of the parameters can reduce catastrophic forgetting efficiently in continual learning, but this needs complex tuning to balance the growing number of parameters and barely share any information across tasks. As a result, they struggle to scale to a large number of tasks without significant overhead. In this paper, we validate in the computer vision domain a recent solution called Adaptive Distillation of Adapters (ADA), which is developed to perform continual learning using pre-trained Transformers and Adapters on text classification tasks. We empirically demonstrate on different classification tasks that this method maintains a good predictive performance without retraining the model or increasing the number of model parameters over the time. Besides it is significantly faster at inference time compared to the state-of-the-art methods.
35.Deep Neural Networks pruning via the Structured Perspective Regularization ⬇️
In Machine Learning, Artificial Neural Networks (ANNs) are a very powerful tool, broadly used in many applications. Often, the selected (deep) architectures include many layers, and therefore a large amount of parameters, which makes training, storage and inference expensive. This motivated a stream of research about compressing the original networks into smaller ones without excessively sacrificing performances. Among the many proposed compression approaches, one of the most popular is \emph{pruning}, whereby entire elements of the ANN (links, nodes, channels, \ldots) and the corresponding weights are deleted. Since the nature of the problem is inherently combinatorial (what elements to prune and what not), we propose a new pruning method based on Operational Research tools. We start from a natural Mixed-Integer-Programming model for the problem, and we use the Perspective Reformulation technique to strengthen its continuous relaxation. Projecting away the indicator variables from this reformulation yields a new regularization term, which we call the Structured Perspective Regularization, that leads to structured pruning of the initial architecture. We test our method on some ResNet architectures applied to CIFAR-10, CIFAR-100 and ImageNet datasets, obtaining competitive performances w.r.t.~the state of the art for structured pruning.
36.Increasing Confidence in Adversarial Robustness Evaluations ⬇️
Hundreds of defenses have been proposed to make deep neural networks robust against minimal (adversarial) input perturbations. However, only a handful of these defenses held up their claims because correctly evaluating robustness is extremely challenging: Weak attacks often fail to find adversarial examples even if they unknowingly exist, thereby making a vulnerable network look robust. In this paper, we propose a test to identify weak attacks, and thus weak defense evaluations. Our test slightly modifies a neural network to guarantee the existence of an adversarial example for every sample. Consequentially, any correct attack must succeed in breaking this modified network. For eleven out of thirteen previously-published defenses, the original evaluation of the defense fails our test, while stronger attacks that break these defenses pass it. We hope that attack unit tests - such as ours - will be a major component in future robustness evaluations and increase confidence in an empirical field that is currently riddled with skepticism.
37.Information Entropy Initialized Concrete Autoencoder for Optimal Sensor Placement and Reconstruction of Geophysical Fields ⬇️
We propose a new approach to the optimal placement of sensors for the problem of reconstructing geophysical fields from sparse measurements. Our method consists of two stages. In the first stage, we estimate the variability of the physical field as a function of spatial coordinates by approximating its information entropy through the Conditional PixelCNN network. To calculate the entropy, a new ordering of a two-dimensional data array (spiral ordering) is proposed, which makes it possible to obtain the entropy of a physical field simultaneously for several spatial scales. In the second stage, the entropy of the physical field is used to initialize the distribution of optimal sensor locations. This distribution is further optimized with the Concrete Autoencoder architecture with the straight-through gradient estimator and adversarial loss to simultaneously minimize the number of sensors and maximize reconstruction accuracy. Our method scales linearly with data size, unlike commonly used Principal Component Analysis. We demonstrate our method on the two examples: (a) temperature and (b) salinity fields around the Barents Sea and the Svalbard group of islands. For these examples, we compute the reconstruction error of our method and a few baselines. We test our approach against two baselines (1) PCA with QR factorization and (2) climatology. We find out that the obtained optimal sensor locations have clear physical interpretation and correspond to the boundaries between sea currents.
38.Discrete Morse Sandwich: Fast Computation of Persistence Diagrams for Scalar Data -- An Algorithm and A Benchmark ⬇️
This paper introduces an efficient algorithm for persistence diagram computation, given an input piecewise linear scalar field f defined on a d-dimensional simplicial complex K, with
$d \leq 3$ . Our method extends the seminal "PairCells" algorithm by introducing three main accelerations. First, we express this algorithm within the setting of discrete Morse theory, which considerably reduces the number of input simplices to consider. Second, we introduce a stratification approach to the problem, that we call "sandwiching". Specifically, minima-saddle persistence pairs ($D_0(f)$) and saddle-maximum persistence pairs ($D_{d-1}(f)$) are efficiently computed by respectively processing with a Union-Find the unstable sets of 1-saddles and the stable sets of (d-1)-saddles. This fast processing of the dimensions 0 and (d-1) further reduces, and drastically, the number of critical simplices to consider for the computation of$D_1(f)$ , the intermediate layer of the sandwich. Third, we document several performance improvements via shared-memory parallelism. We provide an open-source implementation of our algorithm for reproducibility purposes. We also contribute a reproducible benchmark package, which exploits three-dimensional data from a public repository and compares our algorithm to a variety of publicly available implementations. Extensive experiments indicate that our algorithm improves by two orders of magnitude the time performance of the seminal "PairCells" algorithm it extends. Moreover, it also improves memory footprint and time performance over a selection of 14 competing approaches, with a substantial gain over the fastest available approaches, while producing a strictly identical output. We illustrate the utility of our contributions with an application to the fast and robust extraction of persistent 1-dimensional generators on surfaces, volume data and high-dimensional point clouds.
39.AS-IntroVAE: Adversarial Similarity Distance Makes Robust IntroVAE ⬇️
Recently, introspective models like IntroVAE and S-IntroVAE have excelled in image generation and reconstruction tasks. The principal characteristic of introspective models is the adversarial learning of VAE, where the encoder attempts to distinguish between the real and the fake (i.e., synthesized) images. However, due to the unavailability of an effective metric to evaluate the difference between the real and the fake images, the posterior collapse and the vanishing gradient problem still exist, reducing the fidelity of the synthesized images. In this paper, we propose a new variation of IntroVAE called Adversarial Similarity Distance Introspective Variational Autoencoder (AS-IntroVAE). We theoretically analyze the vanishing gradient problem and construct a new Adversarial Similarity Distance (AS-Distance) using the 2-Wasserstein distance and the kernel trick. With weight annealing on AS-Distance and KL-Divergence, the AS-IntroVAE are able to generate stable and high-quality images. The posterior collapse problem is addressed by making per-batch attempts to transform the image so that it better fits the prior distribution in the latent space. Compared with the per-image approach, this strategy fosters more diverse distributions in the latent space, allowing our model to produce images of great diversity. Comprehensive experiments on benchmark datasets demonstrate the effectiveness of AS-IntroVAE on image generation and reconstruction tasks.
40.Improving Worst Case Visual Localization Coverage via Place-specific Sub-selection in Multi-camera Systems ⬇️
6-DoF visual localization systems utilize principled approaches rooted in 3D geometry to perform accurate camera pose estimation of images to a map. Current techniques use hierarchical pipelines and learned 2D feature extractors to improve scalability and increase performance. However, despite gains in typical recall@0.25m type metrics, these systems still have limited utility for real-world applications like autonomous vehicles because of their
worst' areas of performance - the locations where they provide insufficient recall at a certain required error tolerance. Here we investigate the utility of usingplace specific configurations', where a map is segmented into a number of places, each with its own configuration for modulating the pose estimation step, in this case selecting a camera within a multi-camera system. On the Ford AV benchmark dataset, we demonstrate substantially improved worst-case localization performance compared to using off-the-shelf pipelines - minimizing the percentage of the dataset which has low recall at a certain error tolerance, as well as improved overall localization performance. Our proposed approach is particularly applicable to the crowdsharing model of autonomous vehicle deployment, where a fleet of AVs are regularly traversing a known route.
41.Disentangling Embedding Spaces with Minimal Distributional Assumptions ⬇️
Interest in understanding and factorizing learned embedding spaces is growing. For instance, recent concept-based explanation techniques analyze a machine learning model in terms of interpretable latent components. Such components have to be discovered in the model's embedding space, e.g., through independent component analysis (ICA) or modern disentanglement learning techniques. While these unsupervised approaches offer a sound formal framework, they either require access to a data generating function or impose rigid assumptions on the data distribution, such as independence of components, that are often violated in practice. In this work, we link conceptual explainability for vision models with disentanglement learning and ICA. This enables us to provide first theoretical results on how components can be identified without requiring any distributional assumptions. From these insights, we derive the disjoint attributions (DA) concept discovery method that is applicable to a broader class of problems than current approaches but yet possesses a formal identifiability guarantee. In an extensive comparison against component analysis and over 300 state-of-the-art disentanglement models, DA stably maintains superior performance, even under varying distributions and correlation strengths.
42.GAN-based Super-Resolution and Segmentation of Retinal Layers in Optical coherence tomography Scans ⬇️
In this paper, we design a Generative Adversarial Network (GAN)-based solution for super-resolution and segmentation of optical coherence tomography (OCT) scans of the retinal layers. OCT has been identified as a non-invasive and inexpensive modality of imaging to discover potential biomarkers for the diagnosis and progress determination of neurodegenerative diseases, such as Alzheimer's Disease (AD). Current hypotheses presume the thickness of the retinal layers, which are analyzable within OCT scans, can be effective biomarkers. As a logical first step, this work concentrates on the challenging task of retinal layer segmentation and also super-resolution for higher clarity and accuracy. We propose a GAN-based segmentation model and evaluate incorporating popular networks, namely, U-Net and ResNet, in the GAN architecture with additional blocks of transposed convolution and sub-pixel convolution for the task of upscaling OCT images from low to high resolution by a factor of four. We also incorporate the Dice loss as an additional reconstruction loss term to improve the performance of this joint optimization task. Our best model configuration empirically achieved the Dice coefficient of 0.867 and mIOU of 0.765.
43.POEM: Out-of-Distribution Detection with Posterior Sampling ⬇️
Out-of-distribution (OOD) detection is indispensable for machine learning models deployed in the open world. Recently, the use of an auxiliary outlier dataset during training (also known as outlier exposure) has shown promising performance. As the sample space for potential OOD data can be prohibitively large, sampling informative outliers is essential. In this work, we propose a novel posterior sampling-based outlier mining framework, POEM, which facilitates efficient use of outlier data and promotes learning a compact decision boundary between ID and OOD data for improved detection. We show that POEM establishes state-of-the-art performance on common benchmarks. Compared to the current best method that uses a greedy sampling strategy, POEM improves the relative performance by 42.0% and 24.2% (FPR95) on CIFAR-10 and CIFAR-100, respectively. We further provide theoretical insights on the effectiveness of POEM for OOD detection.
44.Omni-Seg+: A Scale-aware Dynamic Network for Pathological Image Segmentation ⬇️
Comprehensive semantic segmentation on renal pathological images is challenging due to the heterogeneous scales of the objects. For example, on a whole slide image (WSI), the cross-sectional areas of glomeruli can be 64 times larger than that of the peritubular capillaries, making it impractical to segment both objects on the same patch, at the same scale. To handle this scaling issue, prior studies have typically trained multiple segmentation networks in order to match the optimal pixel resolution of heterogeneous tissue types. This multi-network solution is resource-intensive and fails to model the spatial relationship between tissue types. In this paper, we propose the Omni-Seg+ network, a scale-aware dynamic neural network that achieves multi-object (six tissue types) and multi-scale (5X to 40X scale) pathological image segmentation via a single neural network. The contribution of this paper is three-fold: (1) a novel scale-aware controller is proposed to generalize the dynamic neural network from single-scale to multi-scale; (2) semi-supervised consistency regularization of pseudo-labels is introduced to model the inter-scale correlation of unannotated tissue types into a single end-to-end learning paradigm; and (3) superior scale-aware generalization is evidenced by directly applying a model trained on human kidney images to mouse kidney images, without retraining. By learning from ~150,000 human pathological image patches from six tissue types at three different resolutions, our approach achieved superior segmentation performance according to human visual assessment and evaluation of image-omics (i.e., spatial transcriptomics). The official implementation is available at this https URL.
45.Toward an ImageNet Library of Functions for Global Optimization Benchmarking ⬇️
Knowledge of search-landscape features of BlackBox Optimization (BBO) problems offers valuable information in light of the Algorithm Selection and/or Configuration problems. Exploratory Landscape Analysis (ELA) models have gained success in identifying predefined human-derived features and in facilitating portfolio selectors to address those challenges. Unlike ELA approaches, the current study proposes to transform the identification problem into an image recognition problem, with a potential to detect conception-free, machine-driven landscape features. To this end, we introduce the notion of Landscape Images, which enables us to generate imagery instances per a benchmark function, and then target the classification challenge over a diverse generalized dataset of functions. We address it as a supervised multi-class image recognition problem and apply basic artificial neural network models to solve it. The efficacy of our approach is numerically validated on the noise free BBOB and IOHprofiler benchmarking suites. This evident successful learning is another step toward automated feature extraction and local structure deduction of BBO problems. By using this definition of landscape images, and by capitalizing on existing capabilities of image recognition algorithms, we foresee the construction of an ImageNet-like library of functions for training generalized detectors that rely on machine-driven features.
46.Flexible-Rate Learned Hierarchical Bi-Directional Video Compression With Motion Refinement and Frame-Level Bit Allocation ⬇️
This paper presents improvements and novel additions to our recent work on end-to-end optimized hierarchical bi-directional video compression to further advance the state-of-the-art in learned video compression. As an improvement, we combine motion estimation and prediction modules and compress refined residual motion vectors for improved rate-distortion performance. As novel addition, we adapted the gain unit proposed for image compression to flexible-rate video compression in two ways: first, the gain unit enables a single encoder model to operate at multiple rate-distortion operating points; second, we exploit the gain unit to control bit allocation among intra-coded vs. bi-directionally coded frames by fine tuning corresponding models for truly flexible-rate learned video coding. Experimental results demonstrate that we obtain state-of-the-art rate-distortion performance exceeding those of all prior art in learned video coding.
47.Tensor Recovery Based on A Novel Non-convex Function Minimax Logarithmic Concave Penalty Function ⬇️
Non-convex relaxation methods have been widely used in tensor recovery problems, and compared with convex relaxation methods, can achieve better recovery results. In this paper, a new non-convex function, Minimax Logarithmic Concave Penalty (MLCP) function, is proposed, and some of its intrinsic properties are analyzed, among which it is interesting to find that the Logarithmic function is an upper bound of the MLCP function. The proposed function is generalized to tensor cases, yielding tensor MLCP and weighted tensor
$L\gamma$ -norm. Consider that its explicit solution cannot be obtained when applying it directly to the tensor recovery problem. Therefore, the corresponding equivalence theorems to solve such problem are given, namely, tensor equivalent MLCP theorem and equivalent weighted tensor$L\gamma$ -norm theorem. In addition, we propose two EMLCP-based models for classic tensor recovery problems, namely low-rank tensor completion (LRTC) and tensor robust principal component analysis (TRPCA), and design proximal alternate linearization minimization (PALM) algorithms to solve them individually. Furthermore, based on the Kurdyka-Łojasiwicz property, it is proved that the solution sequence of the proposed algorithm has finite length and converges to the critical point globally. Finally, Extensive experiments show that proposed algorithm achieve good results, and it is confirmed that the MLCP function is indeed better than the Logarithmic function in the minimization problem, which is consistent with the analysis of theoretical properties.
48.Deep Learning-Based Defect Classification and Detection in SEM Images ⬇️
This proposes a novel ensemble deep learning-based model to accurately classify, detect and localize different defect categories for aggressive pitches and thin resists (High NA applications).In particular, we train RetinaNet models using different ResNet, VGGNet architectures as backbone and present the comparison between the accuracies of these models and their performance analysis on SEM images with different types of defect patterns such as bridge, break and line collapses. Finally, we propose a preference-based ensemble strategy to combine the output predictions from different models in order to achieve better performance on classification and detection of defects. As CDSEM images inherently contain a significant level of noise, detailed feature information is often shadowed by noise. For certain resist profiles, the challenge is also to differentiate between a microbridge, footing, break, and zones of probable breaks. Therefore, we have applied an unsupervised machine learning model to denoise the SEM images to remove the False-Positive defects and optimize the effect of stochastic noise on structured pixels for better metrology and enhanced defect inspection. We repeated the defect inspection step with the same trained model and performed a comparative analysis for "robustness" and "accuracy" metric with conventional approach for both noisy/denoised image pair. The proposed ensemble method demonstrates improvement of the average precision metric (mAP) of the most difficult defect classes. In this work we have developed a novel robust supervised deep learning training scheme to accurately classify as well as localize different defect types in SEM images with high degree of accuracy. Our proposed approach demonstrates its effectiveness both quantitatively and qualitatively.
49.AI-based computer-aided diagnostic system of chest digital tomography synthesis: Demonstrating comparative advantage with X-ray-based AI systems ⬇️
Compared with chest X-ray (CXR) imaging, which is a single image projected from the front of the patient, chest digital tomosynthesis (CDTS) imaging can be more advantageous for lung lesion detection because it acquires multiple images projected from multiple angles of the patient. Various clinical comparative analysis and verification studies have been reported to demonstrate this, but there were no artificial intelligence (AI)-based comparative analysis studies. Existing AI-based computer-aided detection (CAD) systems for lung lesion diagnosis have been developed mainly based on CXR images; however, CAD-based on CDTS, which uses multi-angle images of patients in various directions, has not been proposed and verified for its usefulness compared to CXR-based counterparts. This study develops/tests a CDTS-based AI CAD system to detect lung lesions to demonstrate performance improvements compared to CXR-based AI CAD. We used multiple projection images as input for the CDTS-based AI model and a single-projection image as input for the CXR-based AI model to fairly compare and evaluate the performance between models. The proposed CDTS-based AI CAD system yielded sensitivities of 0.782 and 0.785 and accuracies of 0.895 and 0.837 for the performance of detecting tuberculosis and pneumonia, respectively, against normal subjects. These results show higher performance than sensitivities of 0.728 and 0.698 and accuracies of 0.874 and 0.826 for detecting tuberculosis and pneumonia through the CXR-based AI CAD, which only uses a single projection image in the frontal direction. We found that CDTS-based AI CAD improved the sensitivity of tuberculosis and pneumonia by 5.4% and 8.7% respectively, compared to CXR-based AI CAD without loss of accuracy. Therefore, we comparatively prove that CDTS-based AI CAD technology can improve performance more than CXR, enhancing the clinical applicability of CDTS.