/BoxYolo

3D-object detection using RGBD images

Primary LanguageC++OtherNOASSERTION

Yolo in Caffe

Instructions

Prepare data

(In this case VOCdevkit)(http://host.robots.ox.ac.uk/pascal/VOC/voc2007/index.html)

cd data/yolo
ln -s /your/path/to/VOCdevkit/ .
[You can also get the data at (https://drive.google.com/drive/folders/0B4JSh9t8oHVmQ09rd3dPYlJOX0U?usp=sharing). Copy this in data/yolo/]
python ./get_list.py
[change related paths in script convert.sh]
./convert.sh

Training

cd examples/yolo
Have a look at the train nets available(Googlenet,ZG,VGG16) and write the name of net you want to use for yolo in gnet_solver
[change related paths in script train.sh]
mkdir models
./train.sh

Testing

cd examples/yolo
./test.sh model_path gpu_id

Reference

You Only Look Once: Unified, Real-Time Object detection(http://arxiv.org/abs/1506.02640)