/i3d_test

Test pre-trained NTU i3d model for action recognition

Primary LanguagePython

Code for Testing I3D model. By Srijan Das (srijan.das@inria.fr)

Requirements:

  1. python 2.7
  2. Keras 2.1.5
  3. Tensorflow 1.4.1
  4. cuda/8.0
  5. cudnn/5.1-cuda-8.0

For help - use python test.py -h

Input - ../data/rgb - path of the video frames (input to the script) ../split/test_${dataset_name}.txt - Videos to be tested

Output - pre-trained models are stored in ../models folder prediction results are stored in results.txt within output folder

For testing -

./job.sh Path to the video_dir example - .job.sh ../data/rgb/NTU/