Train an object detector for $6 by using LLMs as vision interns, says HF's Merve
mervenoyann · x · 2026-11-18
- Merve (HF) argues using frontier LLMs directly for object detection is a "skill issue" — expensive and slow.
- Her alternative: treat LLMs as "vision interns" that auto-label data, then distill a small dedicated detector that costs $6 to train and nearly nothing to run.
- Since agents only pick stale proprietary models and make rookie mistakes, she packaged her favorite keypoint/detection model picks and workflow into a tool — "I distilled myself for you."
More from coding & agent
- Microsoft opens Web IQ to all Azure customers as a search engine for AI agents — JordiRib1 · 2026-11-18
- Dev rebuilds all of Singapore in 1:1 scale with 100+ Astra Codex threads and 1,000+ subagents — gabrielchua · 2026-11-18
- GitHub turned 9,791 AI security findings into 99 verified reports, 55 already fixed — mariorod1 · 2026-11-18
- Coinbase Ventures maps the agentic trading stack: where value accrues when agents trade — davidyin44 · 2026-11-18
- Separate MCP server vs. mounting MCP in FastAPI: does the extra hop matter? — DetectiveNew8385 · 2026-11-18
- Granola Launches an Agent-Facing Version: 1M+ Notes Read by Agents Daily — soleio · 2026-11-18