yolov5-v6.0测速

jupiter
2023-05-28 / 0 评论 / 222 阅读 / 正在检测是否收录...
温馨提示:
本文最后更新于2023年05月28日,已超过272天没有更新,若内容或图片失效,请留言反馈。

1.树莓派4B

yolov5s

(base) pi@raspberrypi:/data/yolov5-6.0 $ python detect.py --source test.mp4 --weight yolov5s.pt
/home/pi/miniconda3/lib/python3.7/site-packages/torchvision/io/image.py:13: UserWarning: Failed to load image Python extension:
  warn(f"Failed to load image Python extension: {e}")
detect: weights=['yolov5s.pt'], source=test.mp4, imgsz=[640, 640], conf_thres=0.25, iou_thres=0.45, max_det=1000, device=, view_img=False, save_txt=False, save_conf=False, save_crop=False, nosave=False, classes=None, agnostic_nms=False, augment=False, visualize=False, update=False, project=runs/detect, name=exp, exist_ok=False, line_thickness=3, hide_labels=False, hide_conf=False, half=False, dnn=False
YOLOv5  2021-10-12 torch 1.12.0 CPU

Fusing layers...
Model Summary: 213 layers, 7225885 parameters, 0 gradients
/home/pi/miniconda3/lib/python3.7/site-packages/torch/functional.py:478: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at  /root/pytorch/aten/src/ATen/native/TensorShape.cpp:2894.)
  return _VF.meshgrid(tensors, **kwargs)  # type: ignore[attr-defined]
video 1/1 (1/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 2 trucks, Done. (0.791s)
video 1/1 (2/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.741s)
video 1/1 (3/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.731s)
video 1/1 (4/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.731s)
video 1/1 (5/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.731s)
video 1/1 (6/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.725s)
video 1/1 (7/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.716s)
video 1/1 (8/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.732s)
video 1/1 (9/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.731s)
video 1/1 (10/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.752s)

yolov5n

(base) pi@raspberrypi:/data/yolov5-6.0 $ python detect.py --source test.mp4 --weight yolov5n.pt
/home/pi/miniconda3/lib/python3.7/site-packages/torchvision/io/image.py:13: UserWarning: Failed to load image Python extension:
  warn(f"Failed to load image Python extension: {e}")
detect: weights=['yolov5n.pt'], source=test.mp4, imgsz=[640, 640], conf_thres=0.25, iou_thres=0.45, max_det=1000, device=, view_img=False, save_txt=False, save_conf=False, save_crop=False, nosave=False, classes=None, agnostic_nms=False, augment=False, visualize=False, update=False, project=runs/detect, name=exp, exist_ok=False, line_thickness=3, hide_labels=False, hide_conf=False, half=False, dnn=False
YOLOv5  2021-10-12 torch 1.12.0 CPU

Fusing layers...
Model Summary: 213 layers, 1867405 parameters, 0 gradients
/home/pi/miniconda3/lib/python3.7/site-packages/torch/functional.py:478: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at  /root/pytorch/aten/src/ATen/native/TensorShape.cpp:2894.)
  return _VF.meshgrid(tensors, **kwargs)  # type: ignore[attr-defined]
video 1/1 (1/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 1 car, 1 truck, Done. (0.390s)
video 1/1 (2/985) /data/yolov5-6.0/test.mp4: 384x640 1 car, 1 truck, Done. (0.379s)
video 1/1 (3/985) /data/yolov5-6.0/test.mp4: 384x640 1 car, 1 truck, Done. (0.358s)
video 1/1 (4/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 1 car, 1 truck, Done. (0.367s)
video 1/1 (5/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 1 car, 1 truck, Done. (0.353s)
video 1/1 (6/985) /data/yolov5-6.0/test.mp4: 384x640 1 car, 1 truck, Done. (0.358s)
video 1/1 (7/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 1 car, 1 truck, Done. (0.352s)
video 1/1 (8/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 1 car, 1 truck, Done. (0.353s)
video 1/1 (9/985) /data/yolov5-6.0/test.mp4: 384x640 1 car, 1 truck, Done. (0.361s)
video 1/1 (10/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 1 car, 1 truck, Done. (0.352s)

2.Jetson AGX Xavier

yolov5s

(base-jupiter) nvidia@xavier:/data/yolov5-6.0$ python detect.py --source test.mp4 --weight yolov5s.pt
detect: weights=['yolov5s.pt'], source=test.mp4, imgsz=[640, 640], conf_thres=0.25, iou_thres=0.45, max_det=1000, device=, view_img=False, save_txt=False, save_conf=False, save_crop=False, nosave=False, classes=None, agnostic_nms=False, augment=False, visualize=False, update=False, project=runs/detect, name=exp, exist_ok=False, line_thickness=3, hide_labels=False, hide_conf=False, half=False, dnn=False
YOLOv5  2021-10-12 torch 1.10.0 CUDA:0 (Xavier, 31920.45703125MB)

Fusing layers...
/home/nvidia/archiconda3/envs/base-jupiter/lib/python3.6/site-packages/torch/functional.py:445: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at  /media/nvidia/NVME/pytorch/pytorch-v1.10.0/aten/src/ATen/native/TensorShape.cpp:2157.)
  return _VF.meshgrid(tensors, **kwargs)  # type: ignore[attr-defined]
Model Summary: 213 layers, 7225885 parameters, 0 gradients, 16.5 GFLOPs
video 1/1 (1/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 2 trucks, Done. (0.071s)
video 1/1 (2/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.064s)
video 1/1 (3/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.064s)
video 1/1 (4/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.063s)
video 1/1 (5/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.064s)
video 1/1 (6/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.063s)
video 1/1 (7/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.064s)
video 1/1 (8/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.064s)
video 1/1 (9/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.063s)
video 1/1 (10/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.063s)
video 1/1 (11/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.064s)
video 1/1 (12/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.063s)
video 1/1 (13/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 2 trucks, Done. (0.064s)
video 1/1 (14/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.064s)
video 1/1 (15/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.064s)
video 1/1 (16/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.063s)
video 1/1 (17/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 2 trucks, Done. (0.064s)
video 1/1 (18/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 2 trucks, Done. (0.063s)
video 1/1 (19/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 2 trucks, Done. (0.064s)
video 1/1 (20/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 1 truck, Done. (0.063s)

yolov5n

(base-jupiter) nvidia@xavier:/data/yolov5-6.0$ python detect.py --source test.mp4 --weight yolov5n.pt
detect: weights=['yolov5n.pt'], source=test.mp4, imgsz=[640, 640], conf_thres=0.25, iou_thres=0.45, max_det=1000, device=, view_img=False, save_txt=False, save_conf=False, save_crop=False, nosave=False, classes=None, agnostic_nms=False, augment=False, visualize=False, update=False, project=runs/detect, name=exp, exist_ok=False, line_thickness=3, hide_labels=False, hide_conf=False, half=False, dnn=False
YOLOv5  2021-10-12 torch 1.10.0 CUDA:0 (Xavier, 31920.45703125MB)

Fusing layers...
/home/nvidia/archiconda3/envs/base-jupiter/lib/python3.6/site-packages/torch/functional.py:445: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at  /media/nvidia/NVME/pytorch/pytorch-v1.10.0/aten/src/ATen/native/TensorShape.cpp:2157.)
  return _VF.meshgrid(tensors, **kwargs)  # type: ignore[attr-defined]
Model Summary: 213 layers, 1867405 parameters, 0 gradients, 4.5 GFLOPs
video 1/1 (1/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 1 car, 1 truck, Done. (0.039s)
video 1/1 (2/985) /data/yolov5-6.0/test.mp4: 384x640 1 car, 1 truck, Done. (0.030s)
video 1/1 (3/985) /data/yolov5-6.0/test.mp4: 384x640 1 car, 1 truck, Done. (0.030s)
video 1/1 (4/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 1 car, 1 truck, Done. (0.030s)
video 1/1 (5/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 1 car, 1 truck, Done. (0.029s)
video 1/1 (6/985) /data/yolov5-6.0/test.mp4: 384x640 1 car, 1 truck, Done. (0.030s)
video 1/1 (7/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 1 car, 1 truck, Done. (0.029s)
video 1/1 (8/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 1 car, 1 truck, Done. (0.030s)
video 1/1 (9/985) /data/yolov5-6.0/test.mp4: 384x640 1 car, 1 truck, Done. (0.030s)
video 1/1 (10/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 1 car, 1 truck, Done. (0.030s)
video 1/1 (11/985) /data/yolov5-6.0/test.mp4: 384x640 1 person, 1 car, 1 truck, Done. (0.029s)
video 1/1 (12/985) /data/yolov5-6.0/test.mp4: 384x640 1 car, 1 truck, Done. (0.030s)
video 1/1 (13/985) /data/yolov5-6.0/test.mp4: 384x640 1 car, 1 bus, 1 truck, Done. (0.030s)
video 1/1 (14/985) /data/yolov5-6.0/test.mp4: 384x640 1 car, 1 bus, 1 truck, Done. (0.029s)
video 1/1 (15/985) /data/yolov5-6.0/test.mp4: 384x640 1 car, 1 bus, 1 truck, Done. (0.030s)
video 1/1 (16/985) /data/yolov5-6.0/test.mp4: 384x640 1 car, 1 bus, 1 truck, Done. (0.030s)
video 1/1 (17/985) /data/yolov5-6.0/test.mp4: 384x640 1 car, 1 bus, 1 truck, Done. (0.030s)
video 1/1 (18/985) /data/yolov5-6.0/test.mp4: 384x640 1 car, 1 bus, 1 truck, Done. (0.029s)
video 1/1 (19/985) /data/yolov5-6.0/test.mp4: 384x640 1 car, 1 bus, 1 truck, Done. (0.030s)
video 1/1 (20/985) /data/yolov5-6.0/test.mp4: 384x640 1 car, 1 bus, 1 truck, Done. (0.030s)

3.Jetson Xavier NX

yolov5s

(base-jupiter) nvidia@nx:/data_jupiter/yolov5-6.0$ python detect.py --source test.mp4 --weight yolov5s.pt
detect: weights=['yolov5s.pt'], source=test.mp4, imgsz=[640, 640], conf_thres=0.25, iou_thres=0.45, max_det=1000, device=, view_img=False, save_txt=False, save_conf=False, save_crop=False, nosave=False, classes=None, agnostic_nms=False, augment=False, visualize=False, update=False, project=runs/detect, name=exp, exist_ok=False, line_thickness=3, hide_labels=False, hide_conf=False, half=False, dnn=False
YOLOv5 2023-5-28 torch 1.10.0 CUDA:0 (Xavier, 7765.4140625MB)

Fusing layers...
/home/nvidia/archiconda3/envs/base-jupiter/lib/python3.6/site-packages/torch/functional.py:445: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at  /media/nvidia/NVME/pytorch/pytorch-v1.10.0/aten/src/ATen/native/TensorShape.cpp:2157.)
  return _VF.meshgrid(tensors, **kwargs)  # type: ignore[attr-defined]
Model Summary: 213 layers, 7225885 parameters, 0 gradients, 16.5 GFLOPs
video 1/1 (1/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 2 trucks, Done. (0.061s)
video 1/1 (2/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.043s)
video 1/1 (3/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.040s)
video 1/1 (4/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.040s)
video 1/1 (5/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.040s)
video 1/1 (6/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.040s)
video 1/1 (7/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.040s)
video 1/1 (8/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.040s)
video 1/1 (9/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.040s)
video 1/1 (10/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.040s)
video 1/1 (11/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.040s)
video 1/1 (12/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.040s)
video 1/1 (13/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 2 trucks, Done. (0.040s)
video 1/1 (14/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.040s)
video 1/1 (15/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.040s)
video 1/1 (16/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 3 trucks, Done. (0.040s)
video 1/1 (17/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 2 trucks, Done. (0.040s)
video 1/1 (18/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 2 trucks, Done. (0.040s)
video 1/1 (19/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 2 trucks, Done. (0.040s)
video 1/1 (20/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 2 airplanes, 1 truck, Done. (0.040s)

yolov5n

(base-jupiter) nvidia@nx:/data_jupiter/yolov5-6.0$ python detect.py --source test.mp4 --weight yolov5n.pt
detect: weights=['yolov5n.pt'], source=test.mp4, imgsz=[640, 640], conf_thres=0.25, iou_thres=0.45, max_det=1000, device=, view_img=False, save_txt=False, save_conf=False, save_crop=False, nosave=False, classes=None, agnostic_nms=False, augment=False, visualize=False, update=False, project=runs/detect, name=exp, exist_ok=False, line_thickness=3, hide_labels=False, hide_conf=False, half=False, dnn=False
YOLOv5 2023-5-28 torch 1.10.0 CUDA:0 (Xavier, 7765.4140625MB)

Fusing layers...
/home/nvidia/archiconda3/envs/base-jupiter/lib/python3.6/site-packages/torch/functional.py:445: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at  /media/nvidia/NVME/pytorch/pytorch-v1.10.0/aten/src/ATen/native/TensorShape.cpp:2157.)
  return _VF.meshgrid(tensors, **kwargs)  # type: ignore[attr-defined]
Model Summary: 213 layers, 1867405 parameters, 0 gradients, 4.5 GFLOPs
video 1/1 (1/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 1 car, 1 truck, Done. (0.049s)
video 1/1 (2/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 car, 1 truck, Done. (0.032s)
video 1/1 (3/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 car, 1 truck, Done. (0.032s)
video 1/1 (4/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 1 car, 1 truck, Done. (0.033s)
video 1/1 (5/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 1 car, 1 truck, Done. (0.032s)
video 1/1 (6/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 car, 1 truck, Done. (0.032s)
video 1/1 (7/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 1 car, 1 truck, Done. (0.032s)
video 1/1 (8/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 1 car, 1 truck, Done. (0.032s)
video 1/1 (9/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 car, 1 truck, Done. (0.032s)
video 1/1 (10/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 1 car, 1 truck, Done. (0.032s)
video 1/1 (11/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 person, 1 car, 1 truck, Done. (0.032s)
video 1/1 (12/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 car, 1 truck, Done. (0.032s)
video 1/1 (13/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 car, 1 bus, 1 truck, Done. (0.032s)
video 1/1 (14/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 car, 1 bus, 1 truck, Done. (0.033s)
video 1/1 (15/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 car, 1 bus, 1 truck, Done. (0.032s)
video 1/1 (16/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 car, 1 bus, 1 truck, Done. (0.033s)
video 1/1 (17/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 car, 1 bus, 1 truck, Done. (0.033s)
video 1/1 (18/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 car, 1 bus, 1 truck, Done. (0.032s)
video 1/1 (19/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 car, 1 bus, 1 truck, Done. (0.032s)
video 1/1 (20/985) /data_jupiter/yolov5-6.0/test.mp4: 384x640 1 car, 1 bus, 1 truck, Done. (0.032s)

4.Jetson Nano

yolov5s

#TODO

yolov5n

#TODO

汇总

设备名Yolov5s测速Yolov5n测速CPUGPU显存成本/价格
树莓派4B731ms/1.37FPS352ms/2.84FPS4 核ARM A72 @ 1.5 GHz约850
Jetson Nano161ms/6.21FPS89ms/11.24FPS4 核ARM A57 @ 1.43 GHz128核Maxwell4GB 64 位 LPDDR4x25.6GB/s约1300
Jetson Xavier NX40ms/25FPS32ms/31.25FPS6 核 NVIDIA Carmel ARM®v8.2 64 位 CPU6MB L2 + 4MB L348 个 Tensor Core+384 个 NVIDIA CUDA Core Volta™ GPU8 GB 128 位 LPDDR4x59.7GB/s约4500
Jetson AGX Xavier64ms/15.63FPS29ms/34.48FPS8 核 NVIDIA Carmel Armv8.2 64 位 CPU8MB L2 + 4MB L364 个 Tensor Core+512 个 NVIDIA CUDA Core Volta™ GPU32GB 256 位 LPDDR4x136.5GB/秒约10000

参考资料

  1. NVIDIA Jetson 嵌入式系统开发者套件和模组
0

评论 (0)

打卡
取消