Could not load rknn model

[复制链接] · 发表于 2020-2-24 16:27:37

Dear admin,
I have a tensorflow model, I can convert it into rknn. However, I could not load that rknn model to run inferene.
The converting code is:

from rknn.api import RKNN



if __name__ == '__main__':



  rknn = RKNN(verbose = True)

  rknn.load_tensorflow(tf_pb='model1.pb',

                        inputs=['input_1'],

                        outputs=['conv2d_24/Sigmoid'],

                        input_size_list=[[256, 256, 1]])

  # rknn.config(channel_mean_value='0 0 0 255', reorder_channel='0 1 2')

  rknn.config(channel_mean_value='0 255', reorder_channel='0 1 2')

  

  rknn.build(do_quantization=True, dataset='./dataset.txt')

  #rknn.export_rknn('./mobilenet-ssd-8.rknn')

  rknn.export_rknn('unet_tf.rknn')

  exit(0)

复制代码

and the loading rknn model and run inference is:

from rknn.api import RKNN

import cv2

if __name__ == '__main__':



  rknn = RKNN(verbose = True)

  print('-->loading model')

  rknn.load_rknn('./unet_tf.rknn')

  #rknn.load_rknn('./yolov3.rknn')

  print('loading model done')



  print('--> Init runtime environment')

  ret = rknn.init_runtime()

  if ret != 0:

          print('Init runtime environment failed')

          exit(ret)

  print('done')

  image = cv2.imread('0test.png')

  print(image.shape)

  gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)

  gray = cv2.resize(gray, (256,256))

  out_boxes2 = rknn.inference(inputs=[gray])

  print(out_boxes2)

复制代码

The error is:

D [print_tensor:136]out(0): id[  85] vtl[1] const[0] shape[ 31, 31, 512, 1   ] fmt[u8 ] qnt[ASM zp=  0, scale=0.015807]

D [setup_node:368]Setup node id[28] uid[48] op[CONCAT]

E [op_check:321]Concat input dims size(31 vs 32)

E [setup_node:383]Check node[28] CONCAT fail

I [remove_client_ops:4942]remove_client_ops num=0

I [remove_client_ops:4953]remove_client_ops finish

Create RKNN model fail!

E Catch exception when init runtime!

E Traceback (most recent call last):

E   File "rknn/api/rknn_base.py", line 988, in rknn.api.rknn_base.RKNNBase.init_runtime

E   File "rknn/api/rknn_runtime.py", line 320, in rknn.api.rknn_runtime.RKNNRuntime.build_graph

E Exception: RKNN init failed. error code: RKNN_ERR_MODEL_INVALID

Init runtime environment failed

复制代码

Clearly, the error is shape missmatched. But why I don't have this error in my tensorflow model? I can run inference normally with my tf model but but not rknn.
The code and models are here:
https://drive.google.com/file/d/ ... JM/view?usp=sharing
Please have a look and give me suggestion. Thanks

只看该作者 · 发表于 2020-2-26 08:27:50

本帖最后由 nvnn 于 2020-2-26 14:27 编辑

hisping 发表于 2020-2-24 18:01
you expect a single channel when transform model，but you input three channels picture when runnin ...

[update] SOLVED!!

Thanks for your reply. From my inference code, I already convert the input into grayscale image, so it should be a single channel input. Please clarify more if I have any mistake.
I've done another experiment that use 3 channels input and is the error still the same. My guess is that the deconvolution layer is not converted correctly in rknn. it misses 1 pixel when using decovolution layer?