Belem Zhang
Belem Zhang
1. Visit https://intel.github.io/webml-polyfill/examples/style_transfer/?prefer=none&b=WASM&m=fast_style_transfer_tflite&s=image&d=0&f=WebNN Actual Result: Wrong inference result with Wasm backend  @BruceDai please check if it's caused by the tf.js upgrade.
It's Q4 IMBO, we must complete it, support INT8 models for Semantic Segmentation and Image Classification in Web AI Workload. @BruceDai @mingmingtasd @lisa0314
We need to optimize Deeplab performance in Q4. @ibelem will integration INT8 models based on @mingmingtasd and @lisa0314 's efforts in October.
End-users must select framework, category, model, backend, preference and iterations one by one then run the workload testing, even when they played for a while and went back. Please add...
The page loading is slow for WebNN and OpenCV.js combined pages like https://intel.github.io/webml-polyfill/examples/image_classification/. I don't think we need to download webml-polyfill for OpenCV example, and we don't need to download...
We should invesitage RNNoise for Noise Suppression use cases. - https://hacks.mozilla.org/2017/09/rnnoise-deep-learning-noise-suppression/ - https://github.com/xiph/rnnoise [#webmachinelearning/webnn/issues/66](https://github.com/webmachinelearning/webnn/issues/66) Evaluate noise suppression models for required ops
- List models - Justification (in internal)
The models are inaccessible from huggingface.co in China, we need to solve this issue. 1. Please check if we can switch the domain from https://huggingface.co to https://hf-mirror.com/ for Chinese users....