Skip to content

Dataset · ThinkingHub

SteelBench VLM industrial safety benchmark

ThinkingHub/PP

A dataset of annotated CCTV clips for evaluating VLMs in industrial safety.

What's inside

Includes densely annotated CCTV clips from an active steel plant with natural degradations (dust, glare, steam, low light). Designed for action recognition, PPE verification, and safety violation detection in English, with a size ranging from 1K to 10K.

What you can do with it

  • Evaluating vision-language models (VLMs) in industrial scenarios
  • Training occupational safety violation and PPE detection systems
  • Action recognition in manufacturing videos
Tasks
Video classification, Image classification, Visual question answering
Languages
English
Size
1K–10K rows
Downloads this month
25.4K
Likes
0

How to load it

from datasets import load_dataset

ds = load_dataset("ThinkingHub/PP")