Develop an object detection model for Artpec8

[MenahemOwlytics]

My task is to build a model for Artpec8 chips that detects specific objects

After many unsuccessful attempts to convert existing models, I decided to build the model from scratch based on examples/tree/addbf72c3ee5d35ca9fecf7c6a59598d1a54f6b8/tensorflow-to-larod-artpec8

This project is for object classification, not object recognition.

Is there anyone who has built a model of object recognition in Tensorflow 1 from scratch?

Is there any source code for building this working TFlite model?

https://github.com/google-coral/edgetpu/raw/master/test_data/ssd_mobilenet_v2_coco_quant_postprocess.tflite

Thanks

[Corallo]

Hi @MenahemOwlytics

I was a bit confused by the name you used.
The example we have is for image classification.
For Object detection we don't have a guide to train, but we can recommend using the tensorflow object detector API.

See here:
https://github.com/google-coral/tutorials/blob/master/retrain_ssdlite_mobiledet_qat_tf1.ipynb
and to convert to tflite:
https://github.com/tensorflow/tensorflow/blob/master/tensorflow/lite/g3doc/r1/convert/cmdline_examples.md