- Notifications
You must be signed in to change notification settings - Fork7
Min-Entropy Latent Model for Weakly Supervised Object Detection
WanFang13/MELM
Folders and files
| Name | Name | Last commit message | Last commit date | |
|---|---|---|---|---|
Repository files navigation
A simplified version of MELM with context in PyTorch is released[here] byvasgaowei.
- Linux (tested on ubuntu 14.04LTS)
- NVIDIA GPU + CUDA CuDNN
- Torch7
Install the dependencies
luarocks install hdf5 matio protobuf rapidjson loadcaffe xml
Download dataset, proposals and ImageNet pre-trained model
Download VOC2007 from:http://host.robots.ox.ac.uk/pascal/VOC/voc2007/VOCtrainval_06-Nov-2007.tarhttp://host.robots.ox.ac.uk/pascal/VOC/voc2007/VOCtest_06-Nov-2007.tar
Download proposals from:https://dl.dropboxusercontent.com/s/orrt7o6bp6ae0tc/selective_search_data.tgz
Download VGGF from:http://www.robots.ox.ac.uk/~vgg/software/deep_eval/releases/bvlc/VGG_CNN_F.caffemodelhttps://gist.githubusercontent.com/ksimonyan/a32c9063ec8e1118221a/raw/6a3b8af023bae65669a4ceccd7331a5e7767aa4e/VGG_CNN_F_deploy.prototxt
The data folder has the following structure:
$MELM/data/datasets/VOCdevkit_2007/$MELM/data/datasets/VOCdevkit_2007/VOCcode$MELM/data/datasets/VOCdevkit_2007/VOC2007$MELM/data/datasets/VOCdevkit_2007/...$MELM/data/datasets/proposals/$MELM/data/models/$MELM/data/results/
Install functions
cd ./MELMexport DIR=$(pwd)cd$DIR/utils/c-cuda-functionssh install.shcd$DIR/layersluarocks make
Train and test
cd$DIRsh Run_MELM.sh 0 VOC2007 VGGF SSW 0.1 None melm
This work would not have been possible without prior work: Vadim Kantorov'scontextlocnet, Spyros Gidaris'sLocNet, Sergey Zagoruyko'sloadcaffe, Facebook FAIR'sfbnn/Optim.lua.