Point the browser at an image — classify what it is, or detect every object in it. The models run right here, on your device.
Or try a sample
Verdict
—
Classify an image to see the verdict.
All predictions
Or try a sample
0 objects detected
Two ideas, one browser: classification names the whole image with one label (MobileNet), while detection finds and boxes each object separately (COCO-SSD). Both models run entirely on your device — nothing is uploaded — and they only recognize the categories they were trained on.
Source on GitHub. Sample photos under Creative Commons — credits.