yolov4-tiny-relu-new_coords-tpu on edgetpu too slow

[alan303138]

I am using yolov4-tiny.weights with yolov4-tiny-relu-new_coords.cfg and using yolov4-tiny-relu-new_coords-tpu.cfg convert to tflite with compile.
But the perform on the edgetpu is 60sec per image.

btw,the yolov4-tiny-relu.cfg and yolov4-tiny-relu-tpu.cfg on edgetpu is 0.18 per img.

Yolov4:v3.1.0
tf:2.4.1

[hhk7734]

Would you share your tflite file?

[alan303138]

Here

yolov4-tiny-relu-int8_edgetpu.zip

[hhk7734]

The new_coords part is added in version 3.2.0.

tensorflow-yolov4/CHANGELOG

Lines 1 to 9 in 3d31292

	tensorflow-yolov4 (3.2.0) unstable; urgency=medium
	* config: add yolov4-tiny-relu-new_coords.cfg
	* c_src: layers: add yolo_tpu_layer_new_coords
	* c_src, common, tf, tflite, mAP: add prob_thresh
	* config: add yolov4-tiny-relu-new_coords-tpu.cfg
	* common: base_class: modify text
	-- Hyeonki Hong <[email protected]> Mon, 22 Feb 2021 01:30:53 +0900

and your model is the below. Perhaps you are not using the -a option.
Ref: #20 (comment)

[hhk7734]

yolov4-tiny-relu-tpu model outputs are 4, x1, logistic(x1), x2, logistic(x2).
and yolov4-tiny-relut-new_coords outputs are 2, logistic(x1), logistic(x2).
So ,in yolov4 3.1.0, if you try to predict new_coords, outputs are 2, but yolov4 try to find 4 outputs, because 3.1.0 don't know new_coords.
Since the computational part is in C, the output difference could probably go beyond the allocated memory and cause problems with the system itself.