|
本帖最后由 jin412 于 2023-9-28 12:07 编辑
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
- [toybrick@toybrick rknn_yolov5_demo]$ ./run_my.sh
- post process config: box_conf_threshold = 0.85, nms_threshold = 0.60
- Loading mode...
- D RKNNAPI: ==============================================
- D RKNNAPI: RKNN VERSION:
- D RKNNAPI: API: 1.7.5 (bb79b30 build: 2023-07-18 16:49:12)
- D RKNNAPI: DRV: 1.7.5 (bb79b30 build: 2023-07-18 10:49:14)
- D RKNNAPI: ==============================================
- 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)
- model input num: 1, output num: 3
- 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
- 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
- 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
- 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
- model is NCHW input fmt
- model input height=640, width=640, channel=3
- once run use 61.091000 ms
- loadLabelName ./model/coco_80_labels_list.txt
- person @ (131 0 195 640) 9.315677
- person @ (81 0 145 640) 7.544122
- person @ (369 299 526 437) 7.534053
- person @ (92 215 131 413) 7.520102
- person @ (0 309 640 319) 7.436389
- person @ (111 0 111 640) 7.031783
- person @ (476 263 633 462) 5.957483
- bus @ (193 251 286 260) 5.915627
- person @ (100 345 379 433) 5.678444
- person @ (433 0 462 640) 5.573804
- bus @ (0 73 640 427) 5.231981
- person @ (199 0 280 640) 5.134318
- person @ (95 200 374 289) 5.099438
- person @ (540 0 569 640) 4.785519
- person @ (0 329 256 417) 4.185585
- person @ (81 0 162 640) 3.529844
- bus @ (316 251 409 260) 3.418228
- person @ (496 32 511 640) 3.107728
- person @ (226 0 242 640) 2.702372
- loop count = 100 , average run 41.097110 ms
- 上面是我生成rknn预测
- 下面是官方预测是正确的
- [toybrick@toybrick rknn_yolov5_demo]$ ./run_demo.sh
- post process config: box_conf_threshold = 0.85, nms_threshold = 0.60
- Loading mode...
- D RKNNAPI: ==============================================
- D RKNNAPI: RKNN VERSION:
- D RKNNAPI: API: 1.7.5 (bb79b30 build: 2023-07-18 16:49:12)
- D RKNNAPI: DRV: 1.7.5 (bb79b30 build: 2023-07-18 10:49:14)
- D RKNNAPI: ==============================================
- 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)
- model input num: 1, output num: 3
- 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
- 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
- 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
- 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
- model is NCHW input fmt
- model input height=640, width=640, channel=3
- once run use 67.280000 ms
- loadLabelName ./model/coco_80_labels_list.txt
- person @ (208 246 287 505) 0.867932
- person @ (481 239 561 526) 0.860113
- person @ (109 234 231 536) 0.860113
- loop count = 100 , average run 33.336930 ms
- [toybrick@toybrick rknn_yolov5_demo]$
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 |
本帖子中包含更多资源
您需要 登录 才可以下载或查看,没有帐号?立即注册
x
|