xiaobai1217/awesome-video-datasets

Video datasets

awesome-video-datasets's Issues

Movie - Condensed Movies Dataset

Hey, awesome repo!

Could I request that a dataset please be added in the "Movie" category

Condensed Movies: Story Based Retrieval with Contextual Embeddings (ACCV 2020)
paper: https://arxiv.org/pdf/2005.04208.pdf
Homepage: https://www.robots.ox.ac.uk/~vgg/research/condensed-movies/
33,976 captioned clips from 3,605 movies, 400K+ face-tracks, 8K+ labelled characters, 20K+ subtitles, densely pre-extracted features for each clip (RGB, Motion, Face, Subtitles, Scene)

Consider adding this RGB-D in-hand manipulation dataset

Here's a Video object segmentation dataset: https://www.rgbdinhandmanipulation.com/

Dataset missing

VSPW: A Large-scale Dataset for Video Scene Parsing in the Wild (CVPR 2021)

404 error

Toyota Smarthome: Real-World Activities of Daily Living

Das, Srijan, et al. "Toyota smarthome: Real-world activities of daily living." Proceedings of the IEEE/CVF International Conference on Computer Vision. 2019.

You may put it under the action reco category.

Fixing EPIC-KITCHENS details

EPIC-KITCHENS has been extended in 2020. The new information (in multiple places in your page) should be:

EPIC-KITCHENS: Scaling Egocentric Vision: The EPIC-KITCHENS Dataset
[https://arxiv.org/abs/2006.13256][Homepage]
100 hours, 37 participants, 29M frames, 90K action segments, 700 variable length videos, 97 verb classes, 300 noun classes, 4053 action classes

Also EPIC-KITCHENS has a video retrieval benchmark and challenge. So it's not only for action recognition.

Thanks for your efforts

Video Retrieval

Hey, this is a great list, good job!

I'd have one comment though: under Video Retrieval you list TRECVID, which is a campaign and not a dataset. TRECVID has several tasks, each using different datasets. Not all of those are publicly available, but some of the larger ones are. The two most relevant here are IACC and V3C. IACC (Internet Archive Creative Commons) is a collection of video from the internet archive which has been used in TRECVID for several years and has in the process accumulated many types of annotations. The V3C (Vimeo Creative Commons Collection) is the newer general purpose video dataset, composed of creative commons content from vimeo and designed to be representative of web video in general. Both of these datasets have papers and resources associated with them. The main ones are https://dl.acm.org/doi/abs/10.1145/1631135.1631141 for IACC and https://link.springer.com/chapter/10.1007/978-3-030-05710-7_29 for V3C (also as preprint https://arxiv.org/abs/1810.04401). V3C is also used for another retrieval campaign, the Video Browser Showdown (https://videobrowsershowdown.org), where IACC was also used in the past. Listing this stuff just under TRECVID is therefore not entirely accurate.

Hi, please add a new language based single object tracking dataset (cvpr 2021)

Wang, X., Shu, X., Zhang, Z., Jiang, B., Wang, Y., Tian, Y., & Wu, F. (2021). Towards More Flexible and Accurate Object Tracking with Natural Language: Algorithms and Benchmark. cvpr 2021.

Paper: https://arxiv.org/pdf/2103.16746
Homepage: https://sites.google.com/view/langtrackbenchmark/

MSU Compressed Video Quality Assessment Dataset (CVQAD)

Hello! I'm impressed by your video datasets collection. Appreciate your work! Could you kindly add the MSU Compressed Video Quality Assessment Dataset (CVQAD)? Links to add: Paper and Homepage.

Hi, please add PoseC3D ("revisiting skeleton-based action recognition") skeleton datasets

This dataset includes state-of-the-art 2D pose estimation results for various datasets: including FineGYM, NTURGB+D, UCF101, and HMDB51, which are available at https://github.com/open-mmlab/mmaction2/tree/master/tools/data/skeleton.

MEVA Dataset

This is a really nice job. I think in activity localization you can add the recent MEVA dataset.

TRECVID video captioning dataset (TRECVID-VTT) and HLVU

Hello,

Thanks for this awesome collection!, and also for adding some of the TRECVID datasets.
Our datasets are either fully available for download OR available after submitting a data agreement form before download.

1. Would you please add the VTT dataset: TRECVID-VTT
   https://ir.nist.gov/tv_vtt_data/
   Paper : https://arxiv.org/abs/2009.09984

2. Also recently we released a small annotated video understanding dataset (HLVU):
   https://drive.google.com/drive/u/1/folders/1q1Ca0aFJrF9tB8hsw-mrI9d4tzy5wlPZ
   Paper : https://dl.acm.org/doi/abs/10.1145/3372278.3390742

Thanks!
George

Dataset inaccessable to the public

UVO: Unidentified Video Objects: A Benchmark for Dense, Open-World Segmentation

Go to their dataset and try to follow the link called Google Drive Link and it says that the organization does not give you permission to see the files.