A #bumblebee -like Livebook smart cell is available in #evision v0.1.22 now! Including

- Face Detection/Recognition
- Text Detection/Recognition
- Image Classification/Segmentation

Enable the smart cell with:
Evision.SmartCell.register_smartcells(Evision.SmartCell.Zoo)

#MyElixirStatus

For anyone wondering about the performance and GPU support:

Some models can achieve realtime performance, like the YuNet one, even on a Raspberry Pi 4.

For local applications, opencv_zoo has some benchmark data for them, please see https://github.com/opencv/opencv_zoo.

However, I haven't found a good way to encode a video stream in Livebook/kino and decode the stream on the server.

For GPU acceleration, currently the precompiled version on x86_64 Linux can support OpenCL (FP32/FP16)

GitHub - opencv/opencv_zoo: Model Zoo For OpenCV DNN and Benchmarks.

Model Zoo For OpenCV DNN and Benchmarks. Contribute to opencv/opencv_zoo development by creating an account on GitHub.

GitHub
I haven't added CUDA support yet, but it will be added with modules from opencv_contrib. And hopefully, I can release multiple precompiled versions with/without CUDA support.
@cocoa I’m not sure if you are aware but the woman in the face detection image has asked people to stop using that image. https://en.m.wikipedia.org/wiki/Lenna
Lenna - Wikipedia

@jbosse Oh, I'm sorry I didn't hear that before. I'll remove the image. :)