gemini-4IMAGE
用框选提示 Gemini 4 Argon 生成新检测框
实测显示 Gemini 4 Argon 支持框选提示做目标检测,正负样本都行,文字提示含糊时更好用。
SkalskiP
@skalskip9221 天前收录原推发布于 2026.10.15

SkalskiP @skalskip92原推发布于 2026.10.15
you can also prompt Gemini 4 Argon with boxes to generate new detections - works with positive and negative examples - useful when text prompt is ambiguous - works across images vlms are so versatile... take a look, just the pawns ↓ more box prompts
这条是引用着另一条写的:「Gemini 4 Argon 视觉能力压过 Astra:检测与文字提取第一」
x.com/skalskip92/status/2110424053152698800采集于 2026.10.15 11:25