我正在尝试Yolo v7，它似乎正在工作，但结果图像上没有对象检测映射，而它应该有。
我读了Github如何在Docker上设置Yolo v7，这里是完整的命令，你应该能够重现我的问题。
git clone https://github.com/WongKinYiu/yolov7
cd yolov7
nvidia-docker run --name yolov7 -it --rm -v "$CWD":/yolov7 --shm-size=64g nvcr.io/nvidia/pytorch:21.08-py3
// on the container
cd /yolov7
python -m pip install virtualenv
python -m virtualenv venv3
. venv3/bin/activate
pip install -r requirements.txt
apt update
apt install -y zip htop screen libgl1-mesa-glx
pip install seaborn thop
python detect.py --weights yolov7.pt --conf 0.25 --img-size 640 --source inference/images/horses.jpg
这是最后一个命令的控制台输出:
# python detect.py --weights yolov7.pt --conf 0.25 --img-size 640 --source inference/images/horses.jpg
Namespace(agnostic_nms=False, augment=False, classes=None, conf_thres=0.25, device='', exist_ok=False, img_size=640, iou_thres=0.45, name='exp', no_trace=False, nosave=False, project='runs/detect', save_conf=False, save_txt=False, source='inference/images/horses.jpg', update=False, view_img=False, weights=['yolov7.pt'])
YOLOR 🚀 v0.1-115-g072f76c torch 1.13.0+cu117 CUDA:0 (NVIDIA GeForce GTX 1650, 3903.875MB)
Fusing layers...
RepConv.fuse_repvgg_block
RepConv.fuse_repvgg_block
RepConv.fuse_repvgg_block
Model Summary: 306 layers, 36905341 parameters, 6652669 gradients
Convert model to Traced-model...
traced_script_module saved!
model is traced!
/yolov7/venv3/lib/python3.8/site-packages/torch/functional.py:504: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at ../aten/src/ATen/native/TensorShape.cpp:3190.)
return _VF.meshgrid(tensors, **kwargs) # type: ignore[attr-defined]
Done. (150.9ms) Inference, (0.3ms) NMS
The image with the result is saved in: runs/detect/exp6/horses.jpg
Done. (0.616s)
现在，我应该能够看到从原始图像inference/images/horses.jpg到生成图像runs/detect/exp6/horses.jpg的检测，对吧?但我看到的两幅图像是一样的，没有区别。这个装置有什么问题吗?
Nvidia司机:
$ nvidia-smi
Tue Dec 6 09:47:03 2022
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 525.60.11 Driver Version: 525.60.11 CUDA Version: 12.0 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|===============================+======================+======================|
| 0 NVIDIA GeForce ... On | 00000000:01:00.0 Off | N/A |
| 45% 27C P8 N/A / 75W | 13MiB / 4096MiB | 0% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=============================================================================|
| 0 N/A N/A 1152 G /usr/lib/xorg/Xorg 9MiB |
| 0 N/A N/A 1256 G /usr/bin/gnome-shell 2MiB |
+-----------------------------------------------------------------------------+
$ lsb_release -a
No LSB modules are available.
Distributor ID: Ubuntu
Description: Ubuntu 20.04.4 LTS
Release: 20.04
Codename: focal
将detect.py文件中的变量half修改为False
第31行:half = False
当我们使用Gpu时。该程序只允许使用模型的一半精度。这是在detect.py的第31行中写的内容(#半精度仅在CUDA上支持)。
将其更改为False:第31行:half = False为我变魔术
我也遇到过同样的问题。基本上就是别人提到的。
原因在第31行half = device.type != 'cpu' # half precision only supported on CUDA.
并非所有GPU或所有Nvidia GPU与CUDA支持半精度(16位)浮点数，特别是如果你的GPU有点老。在我的情况下，我使用的是AMD 5700XT(通过ROCM) -这个GPU也没有fp16支持!
为了使其可配置，我添加了一个命令行参数，该参数允许用户覆盖其他答案中提到的变量half;
# After Line 31~, check if command line override is present.
# device = select_device(opt.device)
half = opt.fp16 and device.type != 'cpu' # half precision only supported on CUDA
# After line (169~) with `parser = argparse.ArgumentParser()`
parser.add_argument("--fp16", type=bool, default=False, help="Use float16 (Some GPUs only)")