Skillnoob

Would love for this to get merged as the posted testing results look very promising for a improved performance. Could be especially useful on devices like Raspberry Pi's or Phones...

@sushreebarsa The model has been converted to tflite with int8 and compiled using the latest version of the edgetpu compiler. The edgetpu runtime is also on the latest version. The...

@LakshmiKalaKadali https://drive.google.com/file/d/1Vua1ujK07FXpMdM34uFltA0AchEvC0lH/view?usp=sharing

@LakshmiKalaKadali the error you got inside the notebook is to be expected as the model relies on the delegate from the [coral usb accelerator](https://coral.ai/docs/accelerator/get-started#1-install-the-edge-tpu-runtime)

@pkgoogle Hi. I've tried the same program previously on a raspberry pi 4B with python 3.9.2 and had the same crash but couldn't run it through gdb as the pi...

@pkgoogle i followed the installation instructions exactly as written on the installation page

Recently tried to run this using tensorflow-aarch64 2.13.1 and 2.15.0 and the crash still occurs

easier way of reproducing the issue: ``` pip install ultralytics yolo export model=yolov8n.pt format=edgetpu ``` Install the edge tpu runtime as explained [here](https://coral.ai/docs/accelerator/get-started/#1-install-the-edge-tpu-runtime) ``` yolo detect val model=yolov8n_saved_model/yolov8n_full_integer_quant_edgetpu.tflite data=coco128.yaml batch=1...

also tried this with every tensorflow version going from the current latest to 2.12.1 on a raspberry pi 4B with python 3.9.2

This is now resolved due to https://github.com/feranick updating the libedgetpu runtime to support newer tflite_runtime versions in https://github.com/google-coral/edgetpu/issues/812 .