WR92 COURSE

03 / CONNECTED STUDIO · THE FINAL PROJECT03 / 综合实验 · 课程项目

See it. Name it.
Reimagine it.
先识别,再理解,
然后重新创作。

Use two real AI capabilities in one place: detect everyday objects in a photo, turn the labels into a prompt you can edit, then optionally create a new image.在同一个工作台中使用两项真实 AI 能力:识别照片中的物体、把类别变成可修改的提示词,再按需生成新图片。

Detection runs locally in your browser. Only your edited text, seed and steps go to the optional ZeroGPU image generator—not the source photo. Free quota and queue limits apply.检测在浏览器本地运行。可选 ZeroGPU 生成只发送文字、随机种子和步数,不上传原照片;仍受免费额度与排队限制。

01Choose & detect选图并识别 02Review results查看检测结果 03Edit & generate修改并生成

STEP 01 / START HERE第一步 / 从这里开始

Choose a photo选择一张照片

Use a tested COCO example or upload your own. The images are processed locally.选择一张 COCO 示例,或上传自己的照片;检测在浏览器本地运行。

COCO validation examplesCOCO 验证集示例

or use your own或者自己上传

RT-DETR is the recommended classroom default.

0.50

Choose an example or upload an image.

STEP 02 / LOOK CLOSELY第二步 / 观察结果

What did AI find?AI 识别出了什么?

Waiting

Your photo and the real detection boxes will appear here.原始照片与真实检测框将在此处显示。

No detections yet.
ObjectConfidenceBounding box
Run the detector to see details.

Only recognized object categories are passed to the third step. Detection cannot tell you the full story or artistic style.下一步只接收识别出的物体类别;检测模型无法理解完整故事或艺术风格。

STEP 03 / MAKE SOMETHING NEW第三步 / 重新创作

From labels to image从物体概念到新图片

Detected concepts appear here after Step 1. Review and correct the prompt before generating.第一步完成后,识别出的概念会自动出现。生成前可检查和修改提示词。

NEW IMAGE / RESULT新生成的图片 / 结果

Your new image will appear here after you choose Generate.点击生成后,新图片会显示在这里。

This creates a fresh concept-based image—not a copy or edit of your photo. Live inference is optional and uses visitor ZeroGPU quota.这里会根据概念生成新图,而不是复制或修改原图。实时生成属可选操作,会消耗访客 ZeroGPU 额度。

Open the generator separately单独打开生成工具