RK3399ProD移植部署yolov5-6.2

我的是rk3399pro
大佬我安装你的步骤生成的rknn模型，能不能帮忙找下原因

1. [toybrick@toybrick rknn_yolov5_demo]$ ./run_my.sh
   
2. post process config: box_conf_threshold = 0.85, nms_threshold = 0.60
   
3. Loading mode...
   
4. D RKNNAPI: ==============================================
   
5. D RKNNAPI: RKNN VERSION:
   
6. D RKNNAPI:   API: 1.7.5 (bb79b30 build: 2023-07-18 16:49:12)
   
7. D RKNNAPI:   DRV: 1.7.5 (bb79b30 build: 2023-07-18 10:49:14)
   
8. D RKNNAPI: ==============================================
   
9. sdk version: 1.7.5 (bb79b30 build: 2023-07-18 16:49:12) driver version: 1.7.5 (bb79b30 build: 2023-07-18 10:49:14)
   
10. model input num: 1, output num: 3
    
11.   index=0, name=images_202, n_dims=4, dims=[1, 3, 640, 640], n_elems=1228800, size=1228800, fmt=NCHW, type=UINT8, qnt_type=AFFINE, zp=0, scale=0.003922
    
12.   index=0, name=Conv_Conv_198/out0_0, n_dims=4, dims=[1, 255, 80, 80], n_elems=1632000, size=1632000, fmt=NCHW, type=UINT8, qnt_type=AFFINE, zp=198, scale=0.093099
    
13.   index=1, name=Conv_Conv_199/out0_1, n_dims=4, dims=[1, 255, 40, 40], n_elems=408000, size=408000, fmt=NCHW, type=UINT8, qnt_type=AFFINE, zp=177, scale=0.086641
    
14.   index=2, name=Conv_Conv_200/out0_2, n_dims=4, dims=[1, 255, 20, 20], n_elems=102000, size=102000, fmt=NCHW, type=UINT8, qnt_type=AFFINE, zp=171, scale=0.083522
    
15. model is NCHW input fmt
    
16. model input height=640, width=640, channel=3
    
17. once run use 61.091000 ms
    
18. loadLabelName ./model/coco_80_labels_list.txt
    
19. person @ (131 0 195 640) 9.315677
    
20. person @ (81 0 145 640) 7.544122
    
21. person @ (369 299 526 437) 7.534053
    
22. person @ (92 215 131 413) 7.520102
    
23. person @ (0 309 640 319) 7.436389
    
24. person @ (111 0 111 640) 7.031783
    
25. person @ (476 263 633 462) 5.957483
    
26. bus @ (193 251 286 260) 5.915627
    
27. person @ (100 345 379 433) 5.678444
    
28. person @ (433 0 462 640) 5.573804
    
29. bus @ (0 73 640 427) 5.231981
    
30. person @ (199 0 280 640) 5.134318
    
31. person @ (95 200 374 289) 5.099438
    
32. person @ (540 0 569 640) 4.785519
    
33. person @ (0 329 256 417) 4.185585
    
34. person @ (81 0 162 640) 3.529844
    
35. bus @ (316 251 409 260) 3.418228
    
36. person @ (496 32 511 640) 3.107728
    
37. person @ (226 0 242 640) 2.702372
    
38. loop count = 100 , average run  41.097110 ms
    
39. 
    
40. 上面是我生成rknn预测
    
41. 
    
42. 下面是官方预测是正确的
    
43. [toybrick@toybrick rknn_yolov5_demo]$ ./run_demo.sh
    
44. post process config: box_conf_threshold = 0.85, nms_threshold = 0.60
    
45. Loading mode...
    
46. D RKNNAPI: ==============================================
    
47. D RKNNAPI: RKNN VERSION:
    
48. D RKNNAPI:   API: 1.7.5 (bb79b30 build: 2023-07-18 16:49:12)
    
49. D RKNNAPI:   DRV: 1.7.5 (bb79b30 build: 2023-07-18 10:49:14)
    
50. D RKNNAPI: ==============================================
    
51. sdk version: 1.7.5 (bb79b30 build: 2023-07-18 16:49:12) driver version: 1.7.5 (bb79b30 build: 2023-07-18 10:49:14)
    
52. model input num: 1, output num: 3
    
53.   index=0, name=x.1_0, n_dims=4, dims=[1, 3, 640, 640], n_elems=1228800, size=1228800, fmt=NCHW, type=UINT8, qnt_type=AFFINE, zp=0, scale=0.003922
    
54.   index=0, name=sigmoid_at_1097_147_out0_152, n_dims=4, dims=[1, 255, 80, 80], n_elems=1632000, size=1632000, fmt=NCHW, type=UINT8, qnt_type=AFFINE, zp=0, scale=0.003858
    
55.   index=1, name=sigmoid_at_1109_149_out0_153, n_dims=4, dims=[1, 255, 40, 40], n_elems=408000, size=408000, fmt=NCHW, type=UINT8, qnt_type=AFFINE, zp=0, scale=0.003919
    
56.   index=2, name=sigmoid_at_1121_151_out0_154, n_dims=4, dims=[1, 255, 20, 20], n_elems=102000, size=102000, fmt=NCHW, type=UINT8, qnt_type=AFFINE, zp=0, scale=0.003916
    
57. model is NCHW input fmt
    
58. model input height=640, width=640, channel=3
    
59. once run use 67.280000 ms
    
60. loadLabelName ./model/coco_80_labels_list.txt
    
61. person @ (208 246 287 505) 0.867932
    
62. person @ (481 239 561 526) 0.860113
    
63. person @ (109 234 231 536) 0.860113
    
64. loop count = 100 , average run  33.336930 ms
    
65. [toybrick@toybrick rknn_yolov5_demo]$
    

复制代码附件是我的rknn文件在，扩展名后面 .zip 删除就是正确的文件名