https://github.com/starsdeep/R2Plus1D-MXNet

R2Plus1D MXNet Implementation
https://github.com/starsdeep/R2Plus1D-MXNet

Last synced: 2 months ago
JSON representation

R2Plus1D MXNet Implementation

• Host: GitHub
• URL: https://github.com/starsdeep/R2Plus1D-MXNet
• Owner: starsdeep
• Created: 2018-07-11T04:37:50.000Z (over 6 years ago)
• Default Branch: master
• Last Pushed: 2018-07-11T04:47:33.000Z (over 6 years ago)
• Last Synced: 2024-08-01T22:41:16.487Z (5 months ago)
• Language: Python
• Size: 9.77 KB
• Stars: 11
• Watchers: 6
• Forks: 2
• Open Issues: 1
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

• Awesome-MXNet - R2Plus1D

README

        
# R2Plus1D-mxnet

R2Plus1D MXNet Implementation

R2Plus1D: [A Closer Look at Spatiotemporal Convolutions for Action Recognition (CVPR 2018)](https://arxiv.org/pdf/1711.11248.pdf)

Caffe2 Implementation: https://github.com/facebookresearch/R2Plus1D

## Dataset

[UCF101](http://crcv.ucf.edu/data/UCF101.php)

## Result

Achieved **92.6%** Accuracy(Clip@1, prediction using only 1 clip) on UCF101 Dataset, which is **1.3% higher than the original Caffe2 model**(Accuracy 91.3%).

## Usage

#### Requirements

 

 * MXNet with GPU support

 * opencv

### Data Preparation

 * Download and extract [UCF101](http://crcv.ucf.edu/data/UCF101.php) dataset to ~/UCF101

 * Download pre-trained model from [Caffe2 Pre-trained model](https://github.com/facebookresearch/R2Plus1D/blob/master/tutorials/models.md) to ~/r2.5d_d34_l32.pkl

  

 

#### Training 

 ```

$  python train.py --gpus 0,1,2,3,4,5,6,7 --pretrained ~/r2.5d_d34_l32.pkl --output ~/r2plus1d_output --batch_per_device 4 --lr 1e-4 

--model_depth 34 --wd 0.005 --num_class 101 --num_epoch 80 

```

#### Testing

Assume the training output directory is ~/r2plus1d_output and the epoch number we want to test is 80.

```

$ python validation.py --gpus 0 --output ~/r2plus1d_output --eval_epoch 80 --batch_per_device 48 --model_prefix test 

```