Deep semantic ranking based hashing

This algorithm is described in Deep Semantic Ranking Based Hashing for Multi-Label Image Retrieval.

Main Classes And Functions

• LocRankCostLayer in cudaconvnet/src/layer.cu implements ranking cost layer

• computeLocRankCost in cudaconvnet/src/layer_kernels.cu computes ranking cost

• computeLocRankGrad in cudaconvnet/src/layer_kernels.cu computes gradients of ranking cost

• retrdata.py implements data providers for training and test

• evaluate_nuswide.py computes ndcg, acg and weighted mAP for evaluation

Please find the original cuda-convnet code / wiki from this link: https://code.google.com/p/cuda-convnet2/