search

zldrobit / onnx_tflite_yolov3

A Conversion tool to convert YOLO v3 Darknet weights to TF Lite model (YOLO v3 PyTorch > ONNX > TensorFlow > TF Lite), and to TensorRT (YOLO v3 Pytorch > ONNX > TensorRT).
GNU General Public License v3.0
69 stars 26 forks source link

Use of Reshape feature #6

Closed pra-dan closed 3 years ago

pra-dan commented 4 years ago

Hi, first of all KUDOS to your awesome work. I can't understand what is the purpose of the Reshape Option feature ? I mean, what does it do ?

zldrobit commented 4 years ago

Thanks. Reshape Option literally is the option of reshape op, but it is never used in my observation. According to the tflite flatbuffer definition, Reshape Option consists of only _newshape. You can see from the following diagram, image two functional inputs of reshape op are given, while new_shape is empty.

You could find more information about tflite's Reshape Option in Can not load tflite model.

pra-dan commented 4 years ago

So, it provides an option to remove the reshape ops from the output model... right ?

zldrobit commented 4 years ago

No. We cannot remove the reshape ops, because the computation graph cannot be changed. The second argument of the inputs serve as the shape argument while Reshape Option(_newshape) is useless.

zldrobit commented 3 years ago

Close due to no recent activity. If you have further question, plz reopen it.