Grounding referring expression in computer vision

This project studies the integration of language and vision in computer vision, focusing on Grounding Referring Expressions utilising the state-of-the-art GroundingDINO model. We address the topic of object identification and segmentation, emphasising zero-shot models’ ability to recognise items...

Full description

Saved in:
Bibliographic Details
Main Author: Yuen, Shaun Chien Wee
Other Authors: Hanwang Zhang
Format: Final Year Project
Language:English
Published: Nanyang Technological University 2024
Subjects:
Online Access:https://hdl.handle.net/10356/174979
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
Description
Summary:This project studies the integration of language and vision in computer vision, focusing on Grounding Referring Expressions utilising the state-of-the-art GroundingDINO model. We address the topic of object identification and segmentation, emphasising zero-shot models’ ability to recognise items outside of their training sets. GroundingDINO, an improvement on the DINO model, is essential to our study, as it has significant capabilities in open-set object detection and natural language processing. The project aims to create a Proof of Concept Demo Application demonstrating GroundingDINO’s practical uses in improving human-computer interactions. Our literature review looks into the evolution of computer vision models and the revolutionary characteristics of GroundingDINO and finds gaps in current research, especially in dynamic situations like real-time video analysis. This contributes to the field by highlighting the potential of GroundingDINO in various industries, from surveillance to autonomous systems, and addresses the need for improved language-based object detection in computer vision.