what files and what hierarchy should have the folder that has the data for training. And what are the parameters that we must provide to the train.py file to function properly when training
Nano is an edge device and is not suitable for training due to its limited storage and bandwidth.
If this is essential for you, TensorFlow object detection API might give you some information:
Please noticed that we have lots of desktop GPU designed for fast training.
A recommended step is to train on a desktop GPU and inference the model on Nano afterward.
can we train the model on Nvidia jetson AGX Xavier
i have run a code in pc for face detection and its detecting, same code i am trying to run on nvidia jetson nano
but face is not detecting. I am not understanding whats the problem…? can you please help me out. moreover i am getting the below warning also.
/usr/local/lib/python3.6/dist-packages/face_recognition/api.py:222: RuntimeWarning: invalid value encountered in less_equal
return list(face_distance[known_face_encodings, face_encoding_to_check) <= tolerance)
Xavier is also an edge device.
You can apply training on it but it won’t be the optimal.
It’s recommended to use desktop GPU for the training job instead.
It looks like you are using dlib library, is it correct?
Please noticed that there is a known bug when using dlib on Jetson platform.
You can fix this issue via rebuilding the library.