SmallBigNet
This repo is the official implementation of our paper "SmallBigNet: Integrating Core and Contextual Views for Video Classification (CVPR2020)".
Citation
@inproceedings{li2020smallbignet,
title={Smallbignet: Integrating core and contextual views for video classification},
author={Li, Xianhang and Wang, Yali and Zhou, Zhipeng and Qiao, Yu},
booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
year={2020}
}
Usage
Data Preparation
First, please follow the mmaction2 to prepare data. Note that our codebase only supports the RGB frames. Thus you may need to decord the video dataset offline and store it in SSD. If you need the Kinetics-400 dataset, please feel free to email me. (Tips: if you want to use video online decode, highly recommend you to use the mmaction2. Our idea is simple so only a few codes need to change in resnet3d.py )
K400 Training Scripts
After you prepare the dataset, edit the parameters in scripts/kinectis.sh.
--half indicates using mix precision
--root_path the path you store the whole dataset(RGB)
--train_list_file the train list file (video_name num_frames label)
--val_list_file the val list file (video_name num_frames label)
--model_name [res50, slowonly50, slowonly50_extra, smallbig50_no_extra,smallbig50_extra, smallbig101_no_extra]
--image_tmpl the format of the name you store the RGB frames like img_{:05d}.jpg
If you have any question about the code and data, please contact us directly.