Installation is similar to caffe. Please contact authors of the paper "Learning temporal embeddings for complex video analysis", ICCV2016 for issues specific to this code.
For general caffe installation instructions: See http://caffe.berkeleyvision.org/installation.html for the latest installation instructions.
Check the issue tracker in case you need help: https://github.com/BVLC/caffe/issues
Download the project and model files, and unzip them in the caffe root directory (videovector)
wget http://vision.stanford.edu/vigneshr_data/ICCV15_models/models.zip
unzip models.zipTo test feature extraction (extract our embedding) on sample data, download sample images and run feature extraction
wget http://vision.stanford.edu/vigneshr_data/ICCV15_models/sample_data.zip
unzip sample_data.zip
./projects/videovec_embedding/feature_extraction_pretrained_mednet.sh