lincaiming / des

Single-Shot Object Detection with Enriched Semantics

---

Results on VOC

System	VOC2007 test mAP	VOC2012 test mAP	VOC2007 test mAP trained with COCO	VOC2012 test mAP trained with COCO
SSD300* (VGG16)	77.2	75.8	81.2	79.3
SSD512* (VGG16)	79.8	78.5	83.2	82.2
DES300 (VGG16)	79.7	77.1	82.7	81.0
DES512 (VGG16)	81.7	80.3	84.3	83.7

Results on COCO

System	0.5:0.95	0.5	0.75
SSD300* (VGG16)	25.1	43.1	25.8
SSD512* (VGG16)	28.8	48.5	30.3
DES300 (VGG16)	28.3	47.3	29.4
DES512 (VGG16)	32.8	53.2	34.6

Citing DES

Please cite DES in your publications if it helps your research:

@inproceedings{zhang2018single,
  title = {Single-Shot Object Detection with Enriched Semantics},
  author = {Zhang, Zhishuai and Qiao, Siyuan and Xie, Cihang and Shen, Wei and Wang, Bo and Yuille, Alan L.},
  booktitle = {CVPR},
  year = {2018}
}

Trained Models

1. VOC07

• DES300x300
• DES512x512
• DES300x300_COCO
• DES512x512_COCO

2. VOC12

• DES300x300
• DES512x512
• DES300x300_COCO
• DES512x512_COCO

3. COCO

• DES300x300
• DES512x512

Installation and Preparation

1. Clone the code.

2. Build the code. Please follow Caffe instruction to install all necessary packages and build it.

# Modify Makefile.config according to your Caffe installation.
cp Makefile.config.example Makefile.config
make -j8
# Make sure to include $CAFFE_ROOT/python to your PYTHONPATH.
make py
make test -j8
# (Optional)
make runtest -j8

3. Download VOC2007 and VOC2012 dataset. By default, we assume the data is stored in $HOME/data/

# Download the data.
cd $HOME/data
wget http://host.robots.ox.ac.uk/pascal/VOC/voc2012/VOCtrainval_11-May-2012.tar
wget http://host.robots.ox.ac.uk/pascal/VOC/voc2007/VOCtrainval_06-Nov-2007.tar
wget http://host.robots.ox.ac.uk/pascal/VOC/voc2007/VOCtest_06-Nov-2007.tar
# Extract the data.
tar -xvf VOCtrainval_11-May-2012.tar
tar -xvf VOCtrainval_06-Nov-2007.tar
tar -xvf VOCtest_06-Nov-2007.tar

4. Create the LMDB file.

cd $CAFFE_ROOT
# Create the trainval.txt, test.txt, and test_name_size.txt in data/VOC0712/
./data/VOC0712/create_list.sh
# You can modify the parameters in create_data.sh if needed.
# It will create lmdb files for trainval and test with encoded original image:
#   - $HOME/data/VOCdevkit/VOC0712/lmdb/VOC0712_trainval_lmdb
#   - $HOME/data/VOCdevkit/VOC0712/lmdb/VOC0712_test_lmdb
# and make soft links at examples/VOC0712/
./data/VOC0712/create_data.sh

Train/Eval

All .sh files in jobs folder are for training or evaluation. For training, use the models in initial_models to initialize weights. For evaluation, use the model links above to initialize weights.