Could not work with custom tensorflow lite models

[martinmmi]

Hello, iam martin.

I have problems to run an selfmade tensorflow lite custom model on the edgetpu.

Its a pretrained model on the coco dataset. I choosed the "ssd-mobilenet-v2-fpnlite-320" model for selfmade training.

I converted the model on this introduction: https://coral.ai/docs/edgetpu/compiler/#download

Everytime when i try to running the model with edgetpu, it runs only on the cpu. When i use a finish edge tpu from google:

https://dl.google.com/coral/canned_models/mobilenet_ssd_v2_coco_quant_postprocess_edgetpu.tflite

its working fine with a framerate between 18 and 20 fps.

What can i do?

I think its an unsupported data typ error problem durring the conversion.

I used this introduction for my custom models: https://github.com/EdjeElectronics/TensorFlow-Lite-Object-Detection-on-Android-and-Raspberry-Pi/blob/master/deploy_guides/Raspberry_Pi_Guide.md

Attached is my complete custom model: custom_model_lite.zip

Thanks so much!