Source code of our TPAMI'21 paper Dual Encoding for Video Retrieval by Text and CVPR'19 paper Dual Encoding for Zero-Example Video Retrieval.

Overview

Dual Encoding for Video Retrieval by Text

Source code of our TPAMI'21 paper Dual Encoding for Video Retrieval by Text and CVPR'19 paper Dual Encoding for Zero-Example Video Retrieval.

image

Table of Contents

Environments

  • Ubuntu 16.04
  • CUDA 10.1
  • Python 3.8
  • PyTorch 1.5.1

We used Anaconda to setup a deep learning workspace that supports PyTorch. Run the following script to install the required packages.

conda create --name ws_dual_py3 python=3.8
conda activate ws_dual_py3
git clone https://github.com/danieljf24/hybrid_space.git
cd hybrid_space
pip install -r requirements.txt
conda deactivate

Dual Encoding on MSRVTT10K

Required Data

Run the following script to download and extract MSR-VTT (msrvtt10k-resnext101_resnet152.tar.gz(4.3G)) dataset and a pre-trained word2vec (vec500flickr30m.tar.gz(3.0G). The data can also be downloaded from Baidu pan (url, password:p3p0) or Google drive (url). For more information about the dataset, please refer to here. The extracted data is placed in $HOME/VisualSearch/.

ROOTPATH=$HOME/VisualSearch
mkdir -p $ROOTPATH && cd $ROOTPATH

# download and extract dataset
wget http://8.210.46.84:8787/msrvtt10k-resnext101_resnet152.tar.gz
tar zxf msrvtt10k-resnext101_resnet152.tar.gz -C $ROOTPATH

# download and extract pre-trained word2vec
wget http://lixirong.net/data/w2vv-tmm2018/word2vec.tar.gz
tar zxf word2vec.tar.gz -C $ROOTPATH

Model Training and Evaluation

Run the following script to train and evaluate Dual Encoding network with hybrid space on the official partition of MSR-VTT. The video features are the concatenation of ResNeXt-101 and ResNet-152 features. The code of video feature extraction we used in the paper is available at here.

conda activate ws_dual_py3
./do_all.sh msrvtt10k hybrid resnext101-resnet152

Running the script will do the following things:

  1. Train Dual Encoding network with hybrid space and select a checkpoint that performs best on the validation set as the final model. Notice that we only save the best-performing checkpoint on the validation set to save disk space.
  2. Evaluate the final model on the test set. Note that the dataset has already included vocabulary and concept annotations. If you would like to generate vocabulary and concepts by yourself, run the script ./do_vocab_concept.sh msrvtt10k 1 $ROOTPATH.

If you would like to train Dual Encoding network with the latent space learning (Conference Version), please run the following scrip:

./do_all.sh msrvtt10k latent resnext101-resnet152 $ROOTPATH

To train the model on the Test1k-Miech partition and Test1k-Yu partition of MSR-VTT, please run the following scrip:

./do_all.sh msrvtt10kmiech hybrid resnext101-resnet152 $ROOTPATH
./do_all.sh msrvtt10kyu hybrid resnext101-resnet152 $ROOTPATH

Evaluation using Provided Checkpoints

The overview of pre-trained checkpoints on MSR-VTT is as follows.

Split Pre-trained Checkpoints
Official msrvtt10k_model_best.pth.tar(264M)
Test1k-Miech msrvtt10kmiech_model_best.pth.tar(267M)
Test1k-Yu msrvtt10kyu_model_best.pth.tar(267M)

Note that if you would like to evaluate using our trained checkpoints, please make sure to use the vocabulary and concept annotations that are provided in the msrvtt10k-resnext101_resnet152.tar.gz.

On the official split

Run the following script to download and evaluate our trained checkpoints on the official split of MSR-VTT. The trained checkpoints can also be downloaded from Baidu pan (url, password:p3p0).

MODELDIR=$HOME/VisualSearch/checkpoints
mkdir -p $MODELDIR

# download trained checkpoints
wegt -P $MODELDIR http://8.210.46.84:8787/checkpoints/msrvtt10k_model_best.pth.tar

# evaluate on the official split of MSR-VTT
CUDA_VISIBLE_DEVICES=0 python tester.py --testCollection msrvtt10k --logger_name $MODELDIR  --checkpoint_name msrvtt10k_model_best.pth.tar

On Test1k-Miech and Test1k-Yu splits

In order to evaluate on Test1k-Miech and Test1k-Yu splits, please run the following script.

MODELDIR=$HOME/VisualSearch/checkpoints

# download trained checkpoints on Test1k-Miech
wegt -P $MODELDIR http://8.210.46.84:8787/checkpoints/msrvtt10kmiech_model_best.pth.tar

# evaluate on Test1k-Miech of MSR-VTT
CUDA_VISIBLE_DEVICES=0 python tester.py --testCollection msrvtt10kmiech --logger_name $MODELDIR  --checkpoint_name msrvtt10kmiech_model_best.pth.tar
MODELDIR=$HOME/VisualSearch/checkpoints

# download trained checkpoints on Test1k-Yu
wegt -P $MODELDIR http://8.210.46.84:8787/checkpoints/msrvtt10kyu_model_best.pth.tar

# evaluate on Test1k-Yu of MSR-VTT
CUDA_VISIBLE_DEVICES=0 python tester.py --testCollection msrvtt10kyu --logger_name $MODELDIR  --checkpoint_name msrvtt10kyu_model_best.pth.tar

Expected Performance

The expected performance of Dual Encoding on MSR-VTT is as follows. Notice that due to random factors in SGD based training, the numbers differ slightly from those reported in the paper.

Split Text-to-Video Retrieval Video-to-Text Retrieval SumR
[email protected] [email protected] [email protected] MedR mAP [email protected] [email protected] [email protected] MedR mAP
Official 11.8 30.6 41.8 17 21.4 21.6 45.9 58.5 7 10.3 210.2
Test1k-Miech 22.7 50.2 63.1 5 35.6 24.7 52.3 64.2 5 37.2 277.2
Test1k-Yu 21.5 48.8 60.2 6 34.0 21.7 49.0 61.4 6 34.6 262.6

Dual Encoding on VATEX

Required Data

Download VATEX dataset (vatex-i3d.tar.gz(3.0G)) and a pre-trained word2vec (vec500flickr30m.tar.gz(3.0G)). The data can also be downloaded from Baidu pan (url, password:p3p0) or Google drive (url). For more information about the dataset, please refer to here. Please extract data into $HOME/VisualSearch/.

Model Training and Evaluation

Run the following script to train and evaluate Dual Encoding network with hybrid space on VATEX.

# download and extract dataset
wget http://8.210.46.84:8787/vatex-i3d.tar.gz
tar zxf vatex-i3d.tar.gz -C $ROOTPATH

./do_all.sh vatex hybrid i3d_kinetics $ROOTPATH

Expected Performance

Run the following script to download and evaluate our trained model (vatex_model_best.pth.tar(230M)) on VATEX.

MODELDIR=$HOME/VisualSearch/checkpoints

# download trained checkpoints
wegt -P $MODELDIR http://8.210.46.84:8787/checkpoints/vatex_model_best.pth.tar

CUDA_VISIBLE_DEVICES=0 python tester.py --testCollection vatex --logger_name $MODELDIR  --checkpoint_name vatex_model_best.pth.tar

The expected performance of Dual Encoding with hybrid space learning on MSR-VTT is as follows.

Split Text-to-Video Retrieval Video-to-Text Retrieval SumR
[email protected] [email protected] [email protected] MedR mAP [email protected] [email protected] [email protected] MedR mAP
VATEX 35.8 72.8 82.9 2 52.0 47.5 76.0 85.3 2 39.1 400.3

Dual Encoding on Ad-hoc Video Search (AVS)

Required Data

The following datasets are used for training, validation and testing: the joint collection of MSR-VTT and TGIF, tv2016train and IACC.3. For more information about these datasets, please refer to here.

Frame-level feature data

Please download the frame-level features from Baidu pan (url, password:qwlc). The filename of feature data are summarized as follows.

Datasets 2048-dim ResNeXt-101 2048-dim ResNet-152
MSR-VTT msrvtt10k_ResNext-101.tar.gz msrvtt10k_ResNet-152.tar.gz
TGIF tgif_ResNext-101.tar.gz tgif_ResNet-152.tar.gz
tv2016train tv2016train_ResNext-101.tar.gz tv2016train_ResNet-152.tar.gz
IACC.3 iacc.3_ResNext-101.tar.gz iacc.3_ResNet-152.tar.gz

Note if you have already download MSR-VTT data we provide above, you need not download msrvtt10k_ResNext-101.tar.gz and msrvtt10k_ResNet-152.tar.gz.

Sentence data

Please download the above data, and run the following scripts to extract them into $HOME/VisualSearch/.

ROOTPATH=$HOME/VisualSearch

# extract ResNext-101
tar zxf tgif_ResNext-101.tar.gz -C $ROOTPATH
tar zxf msrvtt10k_ResNext-101.tar.gz -C $ROOTPATH
tar zxf tv2016train_ResNext-101.tar.gz -C $ROOTPATH
tar zxf iacc.3_ResNext-101.tar.gz -C $ROOTPATH

# extract ResNet-152
tar zxf tgif_ResNet-152.tar.gz -C $ROOTPATH
tar zxf msrvtt10k_ResNet-152.tar -C $ROOTPATH
tar zxf tv2016train_ResNet-152.tar.gz -C $ROOTPATH
tar zxf iacc.3_ResNet-152.tar.gz -C $ROOTPATH

# combine feature of tgif and msrvtt10k
./do_combine_features.sh

Train Dual Encoding model from scratch

ROOTPATH=$HOME/VisualSearch
trainCollection=tgif-msrvtt10k
overwrite=0

# Generate a vocabulary on the training set
./util/do_get_vocab.sh $trainCollection $ROOTPATH $overwrite

# Generate concepts according to video captions
./util/do_get_tags.sh $trainCollection $ROOTPATH $overwrite

# Generate video frame info
visual_feature=resnext101-resnet152
./util/do_get_frameInfo.sh $trainCollection $visual_feature $ROOTPATH $overwrite

# training and testing
./do_all_avs.sh $ROOTPATH

How to run Dual Encoding on other datasets?

Our code supports dataset structure:

  • One-folder structure: train, validation and test subset are stored in a folder.
  • Multiple-folder structure: train, validation and test subset are stored in three folders respectively.

One-folder structure

Store the train, validation and test subset into a folder in the following structure.

${collection}
├── FeatureData
│   └── ${feature_name}
│       ├── feature.bin
│       ├── shape.txt
│       └── id.txt
└── TextData
    └── ${collection}train.caption.txt
    └── ${collection}val.caption.txt
    └── ${collection}test.caption.txt
  • FeatureData: video frame features. Using txt2bin.py to convert video frame feature in the required binary format.
  • ${collection}train.caption.txt: training caption data.
  • ${collection}val.caption.txt: validation caption data.
  • ${collection}test.caption.txt: test caption data. The file structure is as follows, in which the video and sent in the same line are relevant.
video_id_1#1 sentence_1
video_id_1#2 sentence_2
...
video_id_n#1 sentence_k
...

Please run the script to generate vocabulary and concepts:

./util/do_vocab_concept.sh $collection 0 $ROOTPATH

Run the following script to train and evaluate Dual Encoding on your own dataset:

./do_all.sh ${collection} hybrid ${feature_name} ${rootpath}

Multiple-folder structure

Store the training, validation and test subsets into three folders in the following structure respectively.

${subset_name}
├── FeatureData
│   └── ${feature_name}
│       ├── feature.bin
│       ├── shape.txt
│       └── id.txt
└── TextData
    └── ${subset_name}.caption.txt
  • FeatureData: video frame features.
  • ${dsubset_name}.caption.txt: caption data of corresponding subset.

You can run the following script to check whether the data is ready:

./do_format_check.sh ${train_set} ${val_set} ${test_set} ${rootpath} ${feature_name}

where train_set, val_set and test_set indicate the name of training, validation and test set, respectively, ${rootpath} denotes the path where datasets are saved and feature_name is the video frame feature name.

Please run the script to generate vocabulary and concepts:

./util/do_vocab_concept.sh ${train_set} 0 $ROOTPATH

If you pass the format check, use the following script to train and evaluate Dual Encoding on your own dataset:

./do_all_multifolder.sh ${train_set} ${val_set} ${test_set} hybrid ${feature_name} ${rootpath}

References

If you find the package useful, please consider citing our TPAMI'21 or CVPR'19 paper:

@article{dong2021dual,
  title={Dual Encoding for Video Retrieval by Text},
  author={Dong, Jianfeng and Li, Xirong and Xu, Chaoxi and Yang, Xun and Yang, Gang and Wang, Xun and Wang, Meng},
  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
  doi = {10.1109/TPAMI.2021.3059295},
  year={2021}
}
@inproceedings{cvpr2019-dual-dong,
title = {Dual Encoding for Zero-Example Video Retrieval},
author = {Jianfeng Dong and Xirong Li and Chaoxi Xu and Shouling Ji and Yuan He and Gang Yang and Xun Wang},
booktitle = {IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
year = {2019},
}
Reference Code for AAAI-20 paper "Multi-Stage Self-Supervised Learning for Graph Convolutional Networks on Graphs with Few Labels"

Reference Code for AAAI-20 paper "Multi-Stage Self-Supervised Learning for Graph Convolutional Networks on Graphs with Few Labels" Please refer to htt

Ke Sun 1 Feb 14, 2022
Write-ups for the SwissHackingChallenge2021 CTF.

SwissHackingChallenge 2021 : Write-ups This repository contains a collection of my write-ups for challenges solved during the SwissHackingChallenge (S

Julien Béguin 3 Jun 07, 2021
Distilling Knowledge via Knowledge Review, CVPR 2021

ReviewKD Distilling Knowledge via Knowledge Review Pengguang Chen, Shu Liu, Hengshuang Zhao, Jiaya Jia This project provides an implementation for the

DV Lab 194 Dec 28, 2022
PSENet - Shape Robust Text Detection with Progressive Scale Expansion Network.

News Python3 implementations of PSENet [1], PAN [2] and PAN++ [3] are released at https://github.com/whai362/pan_pp.pytorch. [1] W. Wang, E. Xie, X. L

1.1k Dec 24, 2022
The code for “Oriented RepPoints for Aerail Object Detection”

Oriented RepPoints for Aerial Object Detection The code for the implementation of “Oriented RepPoints”, Under review. (arXiv preprint) Introduction Or

WentongLi 207 Dec 24, 2022
Volume Control using OpenCV

Gesture-Volume-Control Volume Control using OpenCV Here i made volume control using Python and OpenCV in which we can control the volume of our laptop

Mudit Sinha 3 Oct 10, 2021
A python program to block out your face

Readme This is a small program I threw together in about 6 hours to block out your face. It probably doesn't work very well, so be warned. By default,

1 Oct 17, 2021
SCOUTER: Slot Attention-based Classifier for Explainable Image Recognition

SCOUTER: Slot Attention-based Classifier for Explainable Image Recognition PDF Abstract Explainable artificial intelligence has been gaining attention

87 Dec 26, 2022
Code for generating synthetic text images as described in "Synthetic Data for Text Localisation in Natural Images", Ankush Gupta, Andrea Vedaldi, Andrew Zisserman, CVPR 2016.

SynthText Code for generating synthetic text images as described in "Synthetic Data for Text Localisation in Natural Images", Ankush Gupta, Andrea Ved

Ankush Gupta 1.8k Dec 28, 2022
The official code for the ICCV-2021 paper "Speech Drives Templates: Co-Speech Gesture Synthesis with Learned Templates".

SpeechDrivesTemplates The official repo for the ICCV-2021 paper "Speech Drives Templates: Co-Speech Gesture Synthesis with Learned Templates". [arxiv

Qian Shenhan 53 Dec 23, 2022
Assignment work with webcam

work with webcam : Press key 1 to use emojy on your face Press key 2 to use lip and eye on your face Press key 3 to checkered your face Press key 4 to

Hanane Kheirandish 2 May 31, 2022
Image augmentation for machine learning experiments.

imgaug This python library helps you with augmenting images for your machine learning projects. It converts a set of input images into a new, much lar

Alexander Jung 13.2k Jan 02, 2023
Dirty, ugly, and hopefully useful OCR of Facebook Papers docs released by Gizmodo

Quick and Dirty OCR of Facebook Papers Gizmodo has been working through the Facebook Papers and releasing the docs that they process and review. As lu

Bill Fitzgerald 2 Oct 28, 2021
A simple python program to record security cam footage by detecting a face and body of a person in the frame.

SecurityCam A simple python program to record security cam footage by detecting a face and body of a person in the frame. This code was created by me,

1 Nov 08, 2021
Sort By Face

Sort-By-Face This is an application with which you can either sort all the pictures by faces from a corpus of photos or retrieve all your photos from

0 Nov 29, 2021
This is a project to detect gestures to zoom in or out, using the real-time distance between the index finger and the thumb. It's based on OpenCV and Mediapipe.

Pinch-zoom This is a python project based on real-time hand-gesture detection, to zoom in or out, using the distance between the index finger and the

Harshit Bhalla 6 Jul 11, 2022
Scale-aware Automatic Augmentation for Object Detection (CVPR 2021)

SA-AutoAug Scale-aware Automatic Augmentation for Object Detection Yukang Chen, Yanwei Li, Tao Kong, Lu Qi, Ruihang Chu, Lei Li, Jiaya Jia [Paper] [Bi

Jia Research Lab 182 Dec 29, 2022
Library used to deskew a scanned document

Deskew //Note: Skew is measured in degrees. Deskewing is a process whereby skew is removed by rotating an image by the same amount as its skew but in

Stéphane Brunner 273 Jan 06, 2023
Augmenting Anchors by the Detector Itself

Augmenting Anchors by the Detector Itself Introduction It is difficult to determine the scale and aspect ratio of anchors for anchor-based object dete

4 Nov 06, 2022
Hand gesture detection project with aweome UI implementation.

an awesome hand gesture detection project for you to be creative! Imagination is the limit to do with this project.

AR Ashraf 39 Sep 26, 2022