WebNN Model Inference Demo

On-device AI inference in the browser

Checking…

API Compatibility

Image Classification

Simulates running MobileNetV2 on uploaded images. Uses a lightweight built-in classifier model.

Model Architecture

Input
224×224×3
→
Conv2D
32 filters
→
DepthwiseConv
3×3
→
Bottleneck
×16
→
GlobalAvgPool
1280
→
Dense
1000 classes

MobileNetV2 — optimized for on-device inference with inverted residuals and linear bottlenecks.