Noise-aware-ABP-Saliency

Learning Noise-Aware Encoder-Decoder from Noisy Labels by Alternating Back-Propagation for Saliency Detection

We provide the pytorch version of our code, and the paddlepaddle version code will be release soon.

Saliency Maps

We provide saliency map on eight saliency dataset, including DUT, DUTS, ECSSD, HKU-IS, MSRA-B, THUR, PASCAL-S and SOD.

Please download our saleincy map from: https://drive.google.com/file/d/1uyVkJcTw-2C60nJs2Czt_QoJ-_Z20YBh/view?usp=sharing

Training Dataset

We use DUTS (http://saliencydetection.net/duts/) as our training dataset, and compute saliency maps using three different conventional saliency detection methods: RBD (Saliency optimization from robust backgrounddetection), MR (aliency detection via graph-based manifold ranking) and GS (eodesic saliency using background priors), and treat them as "noisy label".

Please find the noisy saliency map from: https://drive.google.com/file/d/1S0tAG63xMxOnnBPq5aNfE5dYXTTM6kgJ/view?usp=sharing

Testing Dataset

Please find link below for the eight testing datsset we used, incluing both images and their corresponding ground truth saliency maps. https://drive.google.com/drive/folders/1xak2fPZRPnZyzTyzsK2OmdoLk91O9VuQ?usp=sharing

Performance

We show the E-measure and F-measure curves on six benchmark dataset as below:

Trained Model

Please find our trained model from:

Training the model

Testing the model

Contact

Please contact zjnwpu@gmail.com for further discussion.

Reference

Please cite our paper if necessary.

@inproceedings{Zhang2020UCNet,
  title={Learning Noise-Aware Encoder-Decoder from Noisy Labels by Alternating Back-Propagation for Saliency Detection},
  author={Zhang, Jing and Xie, jianwen and Barnes, Nick},
  booktitle={European Conference on Computer Vision},
  year={2020}
}