Toybrick

标题: 关于tensorflow版本的yolo转rknn模型问题 [打印本页]

---

作者: KirutoCode    时间: 2019-9-29 20:36
标题: 关于tensorflow版本的yolo转rknn模型问题
使用的代码如下：https://github.com/ultralytics/yolov3
转rknn的代码如下

1. 
   
2.     # Create RKNN object
   
3.     rknn = RKNN()
   
4. 
   
5.     # Load tensorflow model
   
6.     print('--> Loading model')
   
7.     a = rknn.load_tensorflow(tf_pb='yolov3_coco.pb',inputs=['conv0'],outputs=['conv_lbbox','conv_mbbox','conv_sbbox'],input_size_list=[[416,416,3]])
   
8.     print('return is ',a)
   
9.     print('done')
   
10. 
    
11.     rknn.config(channel_mean_value='0 0 0 255', reorder_channel='0 1 2', batch_size=1)
    
12. 
    
13.     # Build model
    
14.     print('--> Building model')
    
15.     rknn.build(do_quantization=False, pre_compile=False)
    
16.     print('done')
    

复制代码报错如下

1. return is  -1
   
2. done
   
3. --> Building model
   
4. Traceback (most recent call last):
   
5.   File "rknn_transform_onnx_416x416.py", line 27, in <module>
   
6.     rknn.build(do_quantization=False, pre_compile=False)
   
7.   File "/home/dad/anaconda3/envs/market/lib/python3.6/site-packages/rknn/api/rknn.py", line 163, in build
   
8.     inputs = self.rknn_base.net.get_input_layers()
   
9. AttributeError: 'NoneType' object has no attribute 'get_input_layers'
   

复制代码请问是哪里写的不对吗，tensorflow的pb加载失败了。

---

欢迎光临 Toybrick (https://t.rock-chips.com/)

Powered by Discuz! X3.3