Thanks to yun-liu's help. Created by XuanyiLi, if you have any problem in using it, please contact:xuanyili.edu@gmail.com. The best result of my pytorch model is 0.808 ODS F-score now.
the following are the side outputs and the prediction example
If you find our work useful in your research, please consider citing:
@article{RcfEdgePami2019, author = {Yun Liu and Ming-Ming Cheng and Xiaowei Hu and Jia-Wang Bian and Le Zhang and Xiang Bai and Jinhui Tang}, title = {Richer Convolutional Features for Edge Detection}, year = {2019}, journal= {IEEE Trans. Pattern Anal. Mach. Intell.}, volume={}, number={}, pages={}, doi = {}, }
@inproceedings{RCFEdgeCVPR2017, title={Richer Convolutional Features for Edge Detection}, author={Yun Liu and Ming-Ming Cheng, Xiaowei Hu and K Wang and X Bai}, booktitle={IEEE CVPR}, year={2017}, }
this is the edge version of movie Titanic:
youtube video link
I implement the edge detection model according to the RCF model in pytorch.
the result of my pytorch model will be released in the future
| Method | ODS F-score on BSDS500 dataset | ODS F-score on NYU Depth dataset |
|---|---|---|
| ours | 0.808 | *** |
| Reference[1] | 0.811 | *** |
Install pytorch. The code is tested under 0.4.1 GPU version and Python 3.6 on Ubuntu 16.04. There are also some dependencies for a few Python libraries for data processing and visualizations like cv2 etc. It's highly recommended that you have access to GPUs.
To train a RCF model on BSDS500:
python train_RCF.py
After training, to evaluate:
python evaluate.py (for further work)
Side Note: Hello mingyang, I love you
Our code is released under MIT License (see LICENSE file for details).
- Add support for multi-gpu training for the edge detetion task.
- Improve the performance to 0.806/0.811 in the original paper.
- Add a gpu version of edge-eval code to accelerate the evaluation process.
- Add pami version of RCF.
- To download the pretrained model, please click https://drive.google.com/open?id=1TupHeoBKawrniDka0Hc64m3BG4OKG8nM (This pretrained model is not the best model, just for communicating)
- To download the vgg16 pretrained model which is used for the backbone. please click https://drive.google.com/file/d/1lUhPKKj-BSOH7yQL0mOIavvrUbjydPp5/view?usp=sharing.
[1] Richer Convolutional Features for Edge Detection
[2] HED
[4] ContourNet