Skip to content
Implementation of "Sequence to Sequence – Video to Text"
Branch: master
Clone or download
Latest commit 0d6033f Jan 10, 2016
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
.ipynb_checkpoints test function added Dec 8, 2015
.gitignore first commit Dec 6, 2015
LICENSE Create LICENSE Jan 10, 2016
README.md Update README.md Jan 10, 2016
cnn_util.py first commit Dec 6, 2015
cnn_util.pyc There were problems in second LSTM. Now fixed (hopefully) Dec 13, 2015
download_videos.py Update download_videos.py Dec 13, 2015
model.py ss Dec 13, 2015
model.pyc test function added Dec 8, 2015
plane.jpg ss Dec 13, 2015
preprocessing.py There were problems in second LSTM. Now fixed (hopefully) Dec 13, 2015

README.md

video_to_sequence

Usage

  • First you need to download "Microsoft Video Description Corpus"
  • Set "video_data_path" in download_videos.py accordingly.
  • Download Youtube videos by running "download_videos.py"
  • Secondly, you need to preprocess downloaded videos
  • Set paths in cnn_utils.py and preprocess.py
  • Sample & extract features by running "preprocessing.py"
  • Train: train() in model.py
  • You might need to change the paths in "Global Parameters" area according to your environment
  • Test: test() in model.py

alt tag

License

  • BSD License
You can’t perform that action at this time.