/Action_Recognition_3D_CNN

Primary LanguagePython

Action_Recognition_3D_CNN

Description

This is a code for implementation of 3D Convolutional Neural Network for video classification using Keras(tensorflow backend).

Requirements

  • Python 3.7 (Anaconda)
  • Opencv 4.1.0
  • Keras 2.2.4
  • Tensorflow 1.14
  • CUDA 10.1

Dataset

UCF-101 Dataset : Click here

Dataset Setup

First we need to move all the videos in a folder UCF101

python move_videos.py

Model Architecture

Model Architecture

To run the code

python 3dcnn.py --batch 32 --epoch 50 --videos /home/ayush/Activity_Recognition/3DCNN/UCF101/ --nclass 10 --output 3dcnnresult/ --color True --skip False --depth 15

Some Experimental Results

Experiment 1

  • Num Classes : 101
  • Epochs : 100

Loss Plot Accuracy Plot

Experiment 2

  • Num Classes : 50
  • Epochs : 100

Loss Plot Accuracy Plot

Experiment 3

  • Num Classes : 10
  • Num Epochs : 100

Loss Plot Accuracy Plot

Credits