@bobox惊了!Gemini的视觉能力原来强到这种程度! 中发帖

1~2年前,我当时对漫画图片提取分镜很感兴趣。 
当时折腾了好几个月opencv算法,才达到现在Gemini的效果。
现在发现Gemini3.7 flash One Shot就搞定了!
听说Qwen3.8视觉能力很强,试了下还是差Gemini一个档次。
原图:
[image]
Gemini标记图:
[image]
Qwen标记图:
[2_bordered]
本来还想试下Chatgpt,没想到报错了:
We’re so sorry, but the image we created may violate our content policies. If you think we got it wrong, please retry or edit your prompt.
🤣
提示词:用红色边界给漫画图片的分镜绘制边框,注意分镜很多是不规则的图形