Please can anyone help me to convert my model “best.pt” to .tflite to test it on Edge impulse ? because I converted it but when upload and test it in Edge impulse the results of detection was error.
Hello @S234255 first of all welcome to the Edge Impulse community!
Did you try to convert your Pytorch model (.pt) into ONNX first? And then convert from ONNX to TFLite.
Most of the time the .pt files doesn’t contain the architecture, just the weights, which can’t be reconstructed into a model by itself. We recommend to convert to ONNX first.
Let us know how do you proceed here.
I used the ONNX file but the result is also not accurate. Can you please tell me if my configurations are correct or not? and why the border not just on the object?
it should be Ambulance not car
Checking with other colleagues here, it’s relevant that you select the right output layer when you configure the model settings.
Is the output layer (Yolo V5) matching with the model output layer that you actually used to train your model @S234255 ?
I used YOLOv5n to train my model in python therefore I choose here the output layer (Yolo V5
In addition to that I can created Arduino lib its hidden!!!