Skip to content

Latest commit

 

History

History
28 lines (21 loc) · 962 Bytes

README.md

File metadata and controls

28 lines (21 loc) · 962 Bytes
Raw

Actor-Multi-Scale Context Bidirectional Higher Order Interactive Relation Network for Spatio-Temporal Action Localization

notification

  1. This is the pytorch version implementation of paper 'AMCRNet:Actor-Multi-scale Context Bidirectional higher order Interactive Relation Network for Spatio-Temporal Action Localization',rebuilt from mmaction
  2. Trained models are provided in R50_8x8_k400.pth

Update

  1. We correct the typos in Eq.(5) and Eq.(6) as follows.　　

test

  1. preparing data
    python tools\extract_frames.py --video_dir ava_path --frame_dir saving_dir --num_processes nuber_process
  1. run test script
    cd workspace && sh inference_AMCRNet.sh

TODO

  1. Add pre-trained weights of two-stage: R50_4x16_k400.pth、R101_8x8_k400.pth
  2. Add code of one-stage