I'm configured the yolor_csp.cfg for test, modifying the filters for 30 (num classes(5) + 5 * 3), the number classes 5 and, implicit_mul with 30.
But I'm not have a Precision, Recall and, small mAP. However, while I executing test for yolor-p6 model, working not problems.
The csp.cfg working ? I see that the csp.cfg not have a YoloR layer in final part of file cfg. Foremore, the csp.cfg have a 3 implict_mul, different in comparison with p6.cfg.
My Output:
Model Summary: 529 layers, 52519444 parameters, 52519444 gradients
WARNING: --img-size 412 must be multiple of max stride 64, updating to 448
/home/usp/anaconda3/envs/yoloEnv/lib/python3.8/site-packages/torch/nn/functional.py:718: UserWarning: Named tensors and all their associated APIs are an experimental feature and subject to change. Please do not use them for anything important until they are released as stable. (Triggered internally at /pytorch/c10/core/TensorImpl.h:1156.)
return torch.max_pool2d(input, kernel_size, stride, padding, dilation, ceil_mode)
Scanning labels ../../BRA-Dataset/labels/val.cache3 (363 found, 0 missing, 0 empty, 0 duplicate, for 363 images): 363it [00:00, 16860.82it/s]
Class Images Targets P R mAP@.5 mAP@.5:.95: 50%|██████████████████████████████ | 23/46 [00:02<00:01, 11.95it/s]libpng warning: iCCP: known incorrect sRGB profile
Class Images Targets P R mAP@.5 mAP@.5:.95: 100%|████████████████████████████████████████████████████████████| 46/46 [00:04<00:00, 10.71it/s]
all 363 403 0 0 0.00315 0.000539
Anta 363 84 0 0 0.00095 0.000168
Jaguarundi 363 68 0 0 0.00144 0.000282
LoboGuara 363 82 0 0 0.00157 0.000302
OncaParda 363 101 0 0 0.00474 0.000945
TamanduaBandeira 363 68 0 0 0.00704 0.000998
Speed: 6.5/2.9/9.5 ms inference/NMS/total per 448x448 image at batch-size 8
Results saved to runs/test/yolor_csp_val2
Hi everyone
I'm using the command for test in my model trained.
python3 test.py --data ./BRA-Dataset.yaml --img 412 --batch 8 --device 0 --cfg cfg/yolor_csp.cfg --weights ../../PESOS1/bestYoloR-CSP.pt --name yolor_csp_val --verbose --names data/BRA.names
I'm configured the yolor_csp.cfg for test, modifying the filters for 30 (num classes(5) + 5 * 3), the number classes 5 and, implicit_mul with 30.
But I'm not have a Precision, Recall and, small mAP. However, while I executing test for yolor-p6 model, working not problems.
The csp.cfg working ? I see that the csp.cfg not have a YoloR layer in final part of file cfg. Foremore, the csp.cfg have a 3 implict_mul, different in comparison with p6.cfg.
My Output: Model Summary: 529 layers, 52519444 parameters, 52519444 gradients WARNING: --img-size 412 must be multiple of max stride 64, updating to 448 /home/usp/anaconda3/envs/yoloEnv/lib/python3.8/site-packages/torch/nn/functional.py:718: UserWarning: Named tensors and all their associated APIs are an experimental feature and subject to change. Please do not use them for anything important until they are released as stable. (Triggered internally at /pytorch/c10/core/TensorImpl.h:1156.) return torch.max_pool2d(input, kernel_size, stride, padding, dilation, ceil_mode) Scanning labels ../../BRA-Dataset/labels/val.cache3 (363 found, 0 missing, 0 empty, 0 duplicate, for 363 images): 363it [00:00, 16860.82it/s] Class Images Targets P R mAP@.5 mAP@.5:.95: 50%|██████████████████████████████ | 23/46 [00:02<00:01, 11.95it/s]libpng warning: iCCP: known incorrect sRGB profile Class Images Targets P R mAP@.5 mAP@.5:.95: 100%|████████████████████████████████████████████████████████████| 46/46 [00:04<00:00, 10.71it/s] all 363 403 0 0 0.00315 0.000539 Anta 363 84 0 0 0.00095 0.000168 Jaguarundi 363 68 0 0 0.00144 0.000282 LoboGuara 363 82 0 0 0.00157 0.000302 OncaParda 363 101 0 0 0.00474 0.000945 TamanduaBandeira 363 68 0 0 0.00704 0.000998 Speed: 6.5/2.9/9.5 ms inference/NMS/total per 448x448 image at batch-size 8 Results saved to runs/test/yolor_csp_val2