neptuneland / eco-pytorch Goto Github PK

• Trained models on Kinetics dataset for ECO Lite and C3D are provided but they are not final models.

• Keep watching for updates

• Pre-trained model for 2D-Net is provided by tsn-pytorch.

• Python 3.5.2
• PyTorch 0.4.1
• TorchVison: 0.2.1

git clone https://github.com/mzolfaghari/ECO-pytorch

python gen_dataset_lists.py <ucf101/something> <dataset_frames_root_path>

e.g. python gen_dataset_lists.py something ~/dataset/20bn-something-something-v1/

The dataset should be organized as:
<dataset_frames_root_path>/<video_name>/<frame_images>

1. Download the initialization and trained models:

      sh models/download_models.sh

• If you can not access Google Drive, please download the pretrained models from BaiduYun, and put them in "models" folder.

2. Command for training ECO Lite model:

    ./scripts/run_ECOLite_kinetics.sh local

3. For training C3D network use the following command:

    ./scripts/run_c3dres_kinetics.sh local

4. For finetuning on UCF101 use the following command:

    ./scripts/run_ECOLite_finetune_UCF101.sh local

• If you want to train your model from scratch change the config as following:

    --pretrained_parts scratch

• configurations explained in "opts.py"

1. ECO Full
2. Trained models on other datasets

If you use this code or ideas from the paper for your research, please cite our paper:

@inproceedings{ECO_eccv18,
author={Mohammadreza Zolfaghari and
               Kamaljeet Singh and
               Thomas Brox},
title={{ECO:} Efficient Convolutional Network for Online Video Understanding},	       
booktitle={ECCV},
year={2018}
}

Mohammadreza Zolfaghari, Can Zhang

Questions can also be left as issues in the repository. We will be happy to answer them.