Testing Gemini 4 Argon on court mapping: solid but position errors remain
Roger_M_Taylor · x · 2026-10-18
skalskip92 experiments with Gemini 4 Argon on sports vision tasks: first player re-identification by matching players across two images, then court mapping. Overall results are pretty good, though the model still gets some positions wrong. Implementation walkthrough included — a hands-on test of a frontier multimodal model on real-world vision work.
More from Models
- rakyll: open-source harness engineering is just free distribution — models eat it fast — rakyll · 2026-10-18
- Text-only AGI skeptics: all modalities are needed for general intelligence — Angaisb_ · 2026-10-18
- Blogger claims OpenAI quietly cut usage quotas after resetting them — lxfater · 2026-10-18
- ChatGPT Pro spotted auto-creating subagents, but thinking capped at minutes — thomasahle · 2026-10-18
- DeepSeek-V4.1-Flash-1017 GGUF appears on HF via Unsloth, no official announcement yet — MundanePercentage674 · 2026-10-18
- ChatGPT guardrails flag fully-clothed family photos, frustrating photo restoration user — johnellenbe · 2026-10-18