Caffe

This is a merge of inte caffe at https://github.com/intel/caffe and SSD (Single Shot MultiBox Detector) at https://github.com/weiliu89/caffe/tree/ssd

Note

It only targets for inference/scoring, training part is not merged.

Building

Build procedure is the same as on bvlc-caffe-master branch. Both Make and CMake can be used. When OpenMP is available will be used automatically.

Running

1. Prepare/Download the trained model, for example, download from http://www.cs.unc.edu/%7Ewliu/projects/SSD/models_VGGNet_VOC0712_SSD_300x300.tar.gz

2. Run the SSD example: ./build/examples/ssd/ssd_detect.bin ./deploy.prototxt models/VGGNet/VOC0712/SSD_300x300/VGG_VOC0712_SSD_300x300_iter_60000.caffemodel images.txt

Note: Please use the modified deploy.prototxt to get much better performance.