Activity image-based video retrieval

Cross-modal-retrieval

Our approach is focus on Activity Image-to-Video Retrieval (AIVR) task. The compared methods are state-of-the-art single modality hashing methods, multiple modalities hashing methods and cross-modal retrieval methods.

Single modality hashing methods

Some hashing baselines for image retrieval can be found in https://github.com/willard-yuan/hashing-baseline-for-image-retrieval.

Multiple modalities hashing methods

More details refer to https://github.com/czxxjtu/Hash-Learning.github.io. Some details about hashing methods are in hashing-baseline-for-image-retrieval-master folder.

Cross-modal retrieval methods

The compared cross-modal retrieval methods are according to the paper:

Datasets

THUMOS'14 Dataset:

https://pan.baidu.com/s/1H6c8nh_Hs7gVkhESpxtvAg 提取码：qp26

ActivityNet Dataset:

https://pan.baidu.com/s/1P0jRecEmplCPaTPwFoOpVQ 提取码：pnw9

Bibtex

When using images from our dataset, please cite our paper using the following BibTeX[PDF]：

@article{pba2020,
author    = {Ruicong Xu and Li Niu and Jianfu Zhang and Liqing Zhang},
title     = {A Proposal-based Approach for Activity Image-to-Video Retrieval},
journal   = {AAAI},
year      = {2020}}