Vague2Detect improves object detection with ambiguous prompts

Researchers have developed Vague2Detect, a new pipeline designed to improve the accuracy of open-world object detection models when faced with ambiguous or functional prompts. Unlike existing models like YOLO-World, which struggle with vague language, Vague2Detect integrates a knowledge base and a fine-tuned Sentence-BERT model to better interpret user queries. For prompts not covered by the knowledge base, it utilizes GPT-3.5 Turbo to dynamically expand the knowledge base with new concepts. This approach significantly enhances the Vague Prompt Success Rate (VPSR), achieving 61% on a benchmark of household scenes, and up to 85% with the GPT fallback. AI

IMPACT Enhances the ability of AI systems to understand and respond to nuanced, real-world user instructions.

RANK_REASON The cluster contains a research paper detailing a new method for improving object detection. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• GPT-3.5 Turbo
• Ibrohimjon Muminov
• knowledge base
• Open Images V7
• Sentence-BERT
• Vague2Detect
• YOLO
• YOLO-World

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →