Skip to content

Latest commit

 

History

History
16 lines (16 loc) · 849 Bytes

README.md

File metadata and controls

16 lines (16 loc) · 849 Bytes

Image-to-Hazard: GPT-based Logic Reasoning for Hazard Identification in Construction Site using CCTV Data

This paper is accepted by the 41st International Symposium on Automation and Robotics in Construction (ISARC), Lille, France, June 3 - 7 , 2024. Link to access paper.

Proposed approach

image

Demo

image

Sample results

Pose estimation

image

Convert image coordinate to global coordinate

image

Hazard identification with GPT-based models

image

Acknowledgement

This work is inspired by Image2Paragraph, GPT API, OpenMMLab.