yolov4预训练模型生成的.rknn后，推理出现问题，看提示是输...

toybrick@debian10:~/yolov3_demo$ python3 rknn_picture_tiny.py
-->loading model
loading model done
--> Init runtime environment
I NPUTransfer: Starting NPU Transfer Client, Transfer version 1.9.8 (40e4a8a@2020-01-02T09:16:20)
D RKNNAPI: ==============================================
D RKNNAPI: RKNN VERSION:
D RKNNAPI:   API: 1.3.2 (9eebd73 build: 2020-04-02 15:30:36)
D RKNNAPI:   DRV: 1.3.0 (c4f8c23 build: 2019-11-25 10:39:29)
D RKNNAPI: ==============================================
done
E RKNNAPI: rknn_input_set,  inputs[0].buf wrong, buf = 0x23281160, size = 519168 (min_size = 1108992)!
E Catch exception when setting inputs.
E Traceback (most recent call last):
E   File "rknn/api/rknn_base.py", line 1162, in rknn.api.rknn_base.RKNNBase.inference
E   File "rknn/api/rknn_base.py", line 1077, in rknn.api.rknn_base.RKNNBase.set_inputs
E   File "rknn/api/rknn_runtime.py", line 468, in rknn.api.rknn_runtime.RKNNRuntime.set_inputs
E Exception: Set inputs failed. error code: RKNN_ERR_PARAM_INVALID
Traceback (most recent call last):
  File "rknn_picture_tiny.py", line 224, in <module>
    out_boxes, out_boxes2 = rknn.inference(inputs=[mat])
TypeError: cannot unpack non-iterable NoneType object

哪位同仁成功尝试了其他版本的yolov4模型的推理吗？

jefferyzhang 发表于 2020-8-20 22:06
驱动版本太低

你好，这个驱动版本用你们提供的yolov3训练好的模型是正常推理的，yolov4也是应该支持的吧？

jefferyzhang 发表于 2020-8-20 22:06
驱动版本太低

yolov4预训练模型的rknn生成好了，只是推理提示说输入buf的size过小，小于规定最小的min_size？

jefferyzhang 发表于 2020-8-20 22:06
驱动版本太低

问题解决了，应该不是驱动问题，不过还是感谢。从解决问题的地方来看是参照修改的rknn_transform_yolov4.py里的量化函数rknn.build里的参数dataset没设置对，修改成608x608的就顺利完成了推理函数，不过新问题来了，看着是数组不匹配导致，但已经有思路。
toybrick@debian10:~/yolov3_demo$ python3 rknn_picture_608x608.py
-->loading model
loading model done
--> Init runtime environment
I NPUTransfer: Starting NPU Transfer Client, Transfer version 1.9.8 (40e4a8a@2020-01-02T09:16:20)
D RKNNAPI: ==============================================
D RKNNAPI: RKNN VERSION:
D RKNNAPI:   API: 1.3.2 (9eebd73 build: 2020-04-02 15:30:36)
D RKNNAPI:   DRV: 1.3.0 (c4f8c23 build: 2019-11-25 10:39:29)
D RKNNAPI: ==============================================
done
Traceback (most recent call last):
  File "rknn_picture_608x608.py", line 227, in <module>
    out_boxes = out_boxes.reshape(SPAN, LISTSIZE, GRID0, GRID0)
ValueError: cannot reshape array of size 1472880 into shape (3,85,19,19)