Closed Ritesh1991 closed 5 years ago
@Ritesh1991
We need better document for your question, before done, let me go through the procedure quickly:
all image will be passed into detector
(nodejs for state and web/io operation)
https://github.com/SharpAI/DeepCamera/blob/master/src/detector/index.js#L512
we use celery to build worker for easy maintain and future distributed system
For example, we use embedding
work to get feature of preprocessed images through model inference.
embedding
.Currently, we have three task/worker:
Another question is how to run model on ARM GPU
, we leveraged TVM, for the face recognition, we documented here
After converted, integrated as following:
The model train/test code is from insight face (mxnet): https://github.com/deepinsight/insightface/blob/master/deploy/face_embedding.py
document of compile mxnet into tvm: https://docs.tvm.ai/tutorials/nnvm/from_mxnet.html
That's not a lot of works, lol. Just took months for us to figure out the whole procedure which works best.
We also need better architecture to easily integrate new model :)
thanks for quick reply. You people did great work.I will try to use with different model. lets see what happen. and do you have any video of how to use app ?
thanks, @cici-tan is compiling the video. Will release soon.
Hey @Ritesh1991 you can check this first https://github.com/SharpAI/mobile_app_server/blob/android_porting/README.md our video for app is in process
what should we do if we want to add new model into your platform?