Hugging Face's logo

Datasets:
hasnat79
/
ual_bench

Modalities:
Text
Formats:
webdataset
Size:
1K - 10K
ArXiv:
Libraries:
Datasets
WebDataset
License:
Dataset card Data Studio Files Community
1
ual_bench
File size: 955 Bytes 
017c5a7
 
 
9d1a6c7
db74470
 
 
 
 
 
9d1a6c7
a9b34ce
017c5a7
a9b34ce
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
 
---
license: cc-by-4.0
---

# UAL-Bench: The First Comprehensive Unusual Activity Localization Benchmark

[![arXiv](https://img.shields.io/badge/arXiv-2312.14494-b31b1b.svg)](https://arxiv.org/abs/2410.01180)
![unusual--activity--localization](https://img.shields.io/badge/Task-Unusual%20Activity%20Localization-blue)
[![Project Repo](https://img.shields.io/badge/Project--repo-UAL--Bench-green)](https://github.com/Hasnat79/UAL_Bench)

## UAL Bench Datasets
We introduce UAL-Bench, a comprehensive benchmark for unusual activity localization, featuring three video datasets: UAG-OOPS, UAG-SSBD, UAG-FunQA, and an instruction-tune dataset: OOPS-UAG-Instruct.
## Usage
- Go to Files and versions
- download the tar file
- untar the file. For example: 
```bash
tar -xvf uag_oops.tar
```
- Follow the [project repository](https://github.com/Hasnat79/UAL_Bench) for more details including loading the data in your project, implementing VLM-LLM approach etc.