- Notifications
You must be signed in to change notification settings - Fork6
hfslyc/GCPNet
Folders and files
| Name | Name | Last commit message | Last commit date | |
|---|---|---|---|---|
Repository files navigation
This repo is the caffe implementation of the following paper:
Scene Parsing with Global Context Embedding
Wei-Chih Hung,Yi-Hsuan Tsai,Xiaohui Shen,Zhe Lin,Kalyan Sunkavalli,Xin Lu,andMing-Hsuan Yang. In ICCV, 2017.
Please cite our paper if you find it useful for your research.
@inproceedings{hung2017scene, title={Scene Parsing With Global Context Embedding}, author={Hung, Wei-Chih and Tsai, Yi-Hsuan and Shen, Xiaohui and Lin, Zhe and Sunkavalli, Kalyan and Lu, Xin and Yang, Ming-Hsuan}, booktitle={IEEE International Conference on Computer Vision (ICCV)}, year={2017}}- DeepLab-v2 caffe. You will need the this updated version[link] for most recent machine setups.
- A GPU with at least 12GB
- Download theADE20k dataset and put it in
data/.
The directories should be like this:
./data/ADE20k/annotations/validation /images/validation- Download pretrained model
bash get_model.sh- Download precomputed context features/priors of ADE20k val set.
bash get_prior.sh- Execute evaluation script:
python eval.py --prototxt prototxt/ade20k_val.prototxt \ --model models/ade20k_full.caffemodel \ --save-dir results/ade20k/val/ \ --gpu 0The result images will be saved atresults/ade20k/val/.