Inference for yolov4-tiny-darknet is giving error

[Abhishek672]

Hey @david8862,
Thanks for such amazing work.

I am having problem doing inference on yolov4tiny using yolo.py. I have converted .weights to .h5 and generated a anchors.txt file.
while running the snippet :

!python yolo.py --model_type=tiny_yolo4_darknet
--weights_path=weights/keras_darknet.h5
--anchors_path=weights/keras_darknet_anchors.txt
--classes_path=configs/classes.txt
--image
--input=img_0001.jpg
--output=result.jpg

I am getting error :
ValueError('This model type is not supported now')
Traceback (most recent call last):
File "yolo.py", line 91, in _generate_model
yolo_model, _ = get_yolo3_model(self.model_type, num_feature_layers, num_anchors, num_classes, input_shape=self.model_input_shape + (3,), model_pruning=self.pruning_model)
File "/content/keras-YOLOv3-model-set/yolo3/model.py", line 228, in get_yolo3_model
raise ValueError('This model type is not supported now')
ValueError: This model type is not supported now

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "yolo.py", line 437, in
main()
File "yolo.py", line 409, in main
yolo = YOLO_np(**vars(args))
File "yolo.py", line 69, in init
self.yolo_model = self._generate_model()
File "yolo.py", line 104, in _generate_model
assert yolo_model.layers[-1].output_shape[-1] ==
UnboundLocalError: local variable 'yolo_model' referenced before assignment.

Is this repository compatible with AlexyAB yolov4-tiny darknet?

Thanks

[david8862]

@Abhishek672 sorry I didn't involve tiny_yolo4_darknet model type in official process support. Maybe you can try validate_yolo.py for inference if you have been able to convert the darknet .cfg and .weights to Keras .h5

[Abhishek672]

So @david8862,
If I choose yolov4-tiny with cspdarknet53 tiny as the backbone, It is not possible to do the training. Do you know any other repositories where this is implemented?

Thanks