Code and models of paper " ECO: Efficient Convolutional Network for Online Video Understanding", ECCV 2018
Clone or download
Latest commit 3dfee3b Nov 29, 2018
Type Name Latest commit message Commit time
Failed to load latest commit information.
caffe_3d Update Aug 13, 2018
data_list Kinetics-list Aug 1, 2018
doc_files gifs May 7, 2018
models_ECO_Full First commit Jul 30, 2018
models_ECO_Lite Update solver.prototxt Jul 31, 2018
scripts Update extract_frames_main.m Sep 14, 2018
LICENSE Initial commit Apr 18, 2018 Update Nov 29, 2018 Download models Jul 30, 2018 Download models Jul 30, 2018

Code and models of paper. " ECO: Efficient Convolutional Network for Online Video Understanding, European Conference on Computer Vision (ECCV), 2018."

By Mohammadreza Zolfaghari, Kamaljeet Singh, Thomas Brox


🐍 PyTorch implementation for ECO paper now is available here. Many thanks to @zhang-can.


  • 2018.9.06: Providing PyTorch implementation
  • 2018.8.01: Scripts for online recognition and video captioning
  • 2018.7.30: Adding codes and models
  • 2018.4.17: Repository for ECO.


This repository will contains all the required models and scripts for the paper ECO: Efficient Convolutional Network for Online Video Understanding.

In this work, we introduce a network architecture that takes long-term content into account and enables fast per-video processing at the same time. The architecture is based on merging long-term content already in the network rather than in a post-hoc fusion. Together with a sampling strategy, which exploits that neighboring frames are largely redundant, this yields high-quality action classification and video captioning at up to 230 videos per second, where each video can consist of a few hundred frames. The approach achieves competitive performance across all datasets while being 10x to 80x faster than state-of-the-art methods.


Action Recognition on UCF101 and HMDB51 Video Captioning on MSVD dataset

Online Video Understanding Results

Model trained on UCF101 dataset Model trained on Something-Something dataset


  1. Requirements for Python
  2. Requirements for Caffe (see: Caffe installation instructions)


Build Caffe

We used the following configurations with cmake:

  • Cuda 8

  • Python 3

  • Google protobuf 3.1

  • Opencv 3.2

    cd $caffe_3d/
    mkdir build && cd build
    cmake .. 
    make && make install


After successfully completing the installation, you are ready to run all the following experiments.

Data list format

    /path_to_video_folder number_of_frames video_label

Our script for creating kinetics data list.


  1. Download the initialization and trained models:


This will download the following models:

  • Initialization models for 2D and 3D networks (bn_inception_kinetics and 112_c3d_resnet_18_kinetics)

  • Pre-trained models of ECO Lite and ECO Full on the following datasets:

    • Kinetics (400)
    • UCF101
    • HMDB51
    • SomethingSomething (v1)

    *We will provide the results and pre-trained models on Kinetics 600 and SomethingSomething V2 soon.

  1. Train ECO Lite on kinetics dataset:

     sh models_ECO_Lite/kinetics/

Different number of segments

Here, we explain how to modify ".prototxt" to train the network with 8 segments.

  1. In the "VideoData" layer set num_segments: 8
  2. For transform_param of "VideoData" layer copy the following mean values 8 times:
    mean_value: [104]
    mean_value: [117]
    mean_value: [123]
  3. In the "r2Dto3D" layer set the shape as shape { dim: -1 dim: 8 dim: 96 dim: 28 dim: 28 }
  4. Set the kernel_size of "global_pool" layer as kernel_size: [2, 7, 7]


  1. Data
  2. Tables and Results
  3. Demo

License and Citation

All code is provided for research purposes only and without any warranty. Any commercial use requires our consent. If you use this code or ideas from the paper for your research, please cite our paper:

author={Mohammadreza Zolfaghari and
               Kamaljeet Singh and
               Thomas Brox},
title={{ECO:} Efficient Convolutional Network for Online Video Understanding},	       


Mohammadreza Zolfaghari

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