# TCNActionRecognition

**Repository Path**: menghello/TCNActionRecognition

## Basic Information

- **Project Name**: TCNActionRecognition
- **Description**: Skeleton based action recognition models with TCN variants for learning interpretable representation.
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2020-02-20
- **Last Updated**: 2020-12-19

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# Res-TCN for NTURGBD Skeletons dataset
Skeleton based action recognition models with TCN variants for learning interpretable representation.

## Articles
BNMW CVPR2017 paper: https://128.84.21.199/abs/1704.04516v1

## Code
- <b>IMPORTANT</b>: Since the original code release, Keras version has updated. New Keras compatible code with a few minor bug fixes is included under updates/ directory.
- train.py: Contains the keras training routine.
- Models.py: Contains different variants of Res-TCN
- process_skeleton.py: Contains routines for reading in NTURGBD skeleton dataset. This reads in the raw files and creates train/test LMDB files.