README.md

Weakly-/Semi-Supervised Learning in Computer Vision

⌛ Keep Collecting

Included tasks: 1) Object Detection, 2) Semantic Segmentation, 3) Instance Segmentaion, 4) Saliency Detection and 5) Action Localization

1. Object Detection

1.1 Class Label Supervision

• Weakly supervised object detection with convex clustering, CVPR 2015.

• Weakly Supervised Deep Detection Networks, CVPR 2016. [code]

  • Perform simultaneously region selection and classification
  • Region score

• Multiple Instance Detection Network with Online Instance Classifier Refinement, CVPR 2017.

• Weakly Supervised Object Localization with Multi-fold Multiple Instance Learning, PAMI 2017.

• W2F: A Weakly-Supervised to Fully-Supervised Framework for Object Detection, CVPR 2018.

  • Using generated pseudo ground-truth to train fully-supervised RPN

• Min-Entropy Latent Model for Weakly Supervised Object Detection, CVPR 2018.

  • Min-entropy is used as a metric to measure the randomness of object localization during learning, as well as serving as a model to learn object locations.

• Weakly Supervised Region Proposal Network and Object Detection, ECCV 2018.

  • Stage 1: Evaluating bjectness score of sliding windows based on low-level features;
  • Stage 2: Proposal refinement based on classifier;
  • Shared convolutional computations.

• TS2C: Tight Box Mining with Surrounding Segmentation Context for Weakly Supervised Object Detection, ECCV 2018.

  • Semantic Segmentation Confidence Map.
  • Compute objectness scores for the regions inside and outside the box.

• You reap what you sow: Using Videos to Generate High Precision Object Proposals for Weakly-supervised Object Detection, CVPR 2019.

• Dissimilarity Coefficient based Weakly Supervised Object Detection, CVPR 2019.

• WSOD2: Learning Bottom-up and Top-down Objectness Distillation for Weakly-supervised Object Detection, arXiv 1909.

• Towards Precise End-to-end Weakly Supervised Object Detection Network, ICCV 2019.

1.2 Class and Count Labels as Supervision

• C-WSL: Count-guided Weakly Supervised Localization, ECCV 2018.
  • per-class object count
  • count-based region selection algorithm

1.3 Applications

• Cross-Domain Weakly-Supervised Object Detection through Progressive Domain Adaptation, CVPR 2018.

2. Object Localization

• Attention Networks for Weakly Supervised Object Localization, BMVC 2016.

  • computes an attention score on each candidate region in the image.
  • candidate regions are combined together with their attention scores to form a whole-image feature vector. This feature vector is used for classifying the image.

• ContextLocNet: Context-Aware Deep Network Models for Weakly Supervised Localization, ECCV 2016.

  • Leverage their surrounding context regions to improve localization.

• Self-produced Guidance for Weakly-supervised Object Localization, ECCV 2018. [code]

• Self-Erasing Network for Integral Object Attention, NeurIPS 2018.

• Attention-based Dropout Layer for Weakly Supervised Object Localization, CVPR 2019(Oral).[code]

• Integral Object Mining via Online Attention Accumulation, ICCV 2019.

• Dual-attention Focused Module for Weakly Supervised Object Localization, arXiv 1909.

3. Semantic Segmentation

3.1 Class Label Supervision

• Seed, Expand and Constrain: Three Principles for Weakly-Supervised Image Segmentation, ECCV 2016.

• Object Region Mining with Adversarial Erasing: A Simple Classification to Semantic Segmentation Approach, CVPR 2017.

• Exploiting saliency for object segmentation from image level labels, CVPR 2017. [code]

• Two-Phase Learning for Weakly Supervised Object Localization, ICCV 2017.

• Self-Supervised Difference Detection for Weakly-Supervised Semantic Segmentation, ICCV 2019.

• Causal Intervention for Weakly-Supervised Semantic Segmentation, NeurIPS 2020.

3.2 Scribble Supervision

• ScribbleSup: Scribble-Supervised Convolutional Networks for Semantic Segmentation, CVPR 2016.

• On Regularized Losses for Weakly-supervised CNN Segmentation, ECCV 2018.

3.3 Semi-supervised Learning

• Simple Does It: Weakly Supervised Instance and Semantic Segmentation, CVPR 2017. [code]

• Revisiting CycleGAN for semi-supervised segmentation, arXiv 1908.

• Exploiting Temporality for Semi-Supervised Video Segmentation, arXiv 1908.

• Semi-Supervised Semantic Image Segmentation with Self-correcting Networks, CVPR 2020.

4. Instance Segmentation

• PRM: Weakly Supervised Instance Segmentation using Class Peak Response, CVPR 2018.[code]

• IRN: Weakly Supervised Learning of Instance Segmentation with Inter-pixel Relations, CVPR 2019(Oral).[code]

• Object Counting and Instance Segmentation with Image-level Supervision, CVPR 2019.

  • Instance level segmentation using peak response and count knowledge
  • Only per-class counts in subitizing range are provided
  • Baseline: PRM

• Where are the Masks: Instance Segmentation with Image-level Supervision, BMVC 2019.

  • Baseline: IRN

• Weakly Supervised Instance Segmentation by Deep Multi-Task Community Learning, arXiv 2001.

• Label-PEnet: Sequential Label Propagation and Enhancement Networks for Weakly Supervised Instance Segmentation, ICCV 2019.

4. Salient Object Detection

4.1 Image

• Learning to Detect Salient Objects with Image-level Supervision, CVPR 2017. [code]

• Weakly-supervised Salient Object detection using Image Labels, AAAI 2018.

• Multi-source Weak Supervision for saliencly detection, CVPR 2019. [code]

• Joint learning of saliency detection and weakly supervised semantic segmentation, ICCV 2019.

4.2 Video

• Semi-Supervised Video Salient Object Detection Using Pseudo-Labels, ICCV 2019.
  • Generating pixel-wsie pseudo-labels from sparsely annotated frames by using spatial and temporal information.

5. Salient Instance Detection

• Weakly-supervised Salient Instance Detection, BMVC 2020 (Oral, Best student paper runner-up).

6. Action Localization

• Weakly Supervised Energy-Based Learning for Action Segmentation, ICCV 2019 (Oral).

• 3C-Net: Category Count and Center Loss for Weakly-Supervised Action Localization, arXiv 1908.

7. Derain

• Semi-supervised Transfer Learning for Image Rain Removal, CVPR 2019.

8. Useful Slides/Blogs for Comprehensive Reading

• Weakly Supervised Object Localization, by Zongwei Zhou, 2019.

• Towards Weakly Supervised Object Segmentation & Scene Parsing, by Yunchao Wei, Valse Workshop, 2019.

• Weakly Supervised Semantic Segmentation, by Yunchao Wei, Valse Tutorial, 2019.

• Weakly Supervised Object Detection, Localization, and instance segmentation, by Qixiang Ye, 2019.