Web AI Workload
Framework
WebNN OpenCV.js
Category
Image Classification Object Detection Semantic Segmentation Skeleton Detection Super Resolution Emotion Analysis Facial Landmark Detection
Model
MobileNet v1 (TFLite) MobileNet v1 Quant (TFLite) MobileNet v2 (TFLite) MobileNet v2 Quant (TFLite) SqueezeNet (TFLite) Inception v3 (TFLite) Inception v3 Quant (TFLite) Inception v4 (TFLite) Inception v4 Quant (TFLite) Inception ResNet v2 (TFLite) SqueezeNet (ONNX) MobileNet v2 (ONNX) ResNet50 v1 (ONNX) ResNet50 v2 (ONNX) Inception v2 (ONNX) DenseNet 121 (ONNX) SqueezeNet (OpenVino) MobileNet v1 (OpenVino) MobileNet v1 Quant (OpenVino) MobileNet v2 (OpenVino) MobileNet v2 Quant (OpenVino) ResNet50 v1 (OpenVino) ResNet50 v1 Quant (OpenVino) DenseNet 121 (OpenVino) Inception v2 (OpenVino) Inception v4 (OpenVino) Squeezenet (Caffe2) MobileNet v1 Quant (Caffe2)
Backend
WASM
WebGL
WebNN
WASM
SIMD
Threads
Threads+SIMD
Preference
SUSTAINED_SPEED
FAST_SINGLE_ANSWER
LOW_POWER
UTRAL_LOW_POWER
NONE
Iterations
1(warming up) +
Steps to test OpenCV Workload
1. Quit any running instance of Chrome
2. Run your favorite terminal emulator
3. In the terminal, run command: google-chrome --enable-features=SharedArrayBuffer
4. Visit https://intel.github.io/webml-polyfill/workload
5. Run the OpenCV.js workload
Run
(It will take several minutes)