Code release of "Deep Visual-Semantic Quantization of Efficient Image Retrieval" (CVPR 17)
Switch branches/tags
Nothing to show
Clone or download
Pull request Compare This branch is 1 commit behind caoyue10:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.


This is the Tensorflow (Version 0.11) implementation of CVPR-17 paper "Deep Visual-Semantic Quantization for Efficient Image Retrieval". The descriptions of files in this directory are listed below:

  • contains the main implementation (network structure, loss function, optimization procedure and etc.) of the proposed approach dvsq.
  • contains the implementation of dvsq for evaluation.
  • contains the implementation of Dataset, MAP and ProcessBar.
  • gives an example to show how to train dvsq model.
  • gives an example to show how to evaluate the trained quantization model.
  • gives an example to show the full procedure of training and evaluating the proposed approach dvsq.

Data Preparation

In data/nuswide_81/train.txt, we give an example to show how to prepare image training data. In data/nuswide_81/test.txt and data/nuswide_81/database.txt, the list of testing and database images could be processed during predicting procedure. In data/nuswide_81/nuswide_wordvec.txt, we have already prepared the word vectors of the labels extracted by Word2Vec model pretrained on Google News Dataset.

Training Model and Predicting

The bvlc_reference_caffenet is used as the pre-trained model. If the NUS_WIDE dataset and pre-trained caffemodel is prepared, the example can be run with the following command:



  author    = {Yue Cao and
               Mingsheng Long and
               Jianmin Wang and
               Shichen Liu},
  title     = {Deep Visual-Semantic Quantization for Efficient Image Retrieval},
  booktitle = {2017 {IEEE} Conference on Computer Vision and Pattern Recognition,
      {CVPR} 2017, Honolulu, Hawaii, USA, July 21-26, 2017}