参考博客:openpose环境搭建(全教程),
以及利用软连接完美解决由于CUDA版本不匹配造成的各种坑,
Caffe的安装 :官网
CPU版本 不可行,太慢效果也不好
GPU
ubuntu18.04 cuda10 caffe openpose环境配置与使用
Ubuntu20.04 NVIDIA-SMI 450.66 Driver Version: 450.66 CUDA Version: 11.0 CUDNN Version: 8.0.2.39
失败。。。啊啊啊啊 win10,找机会Ubuntu系统再试一次
win10 openpose
VS2019+Python开发
[INFO]Pose estimation.
[INFO]Model loads time: 0.6643552780151367
OpenCV(ocl4dnn): consider to specify kernel configuration cache directory
via OPENCV_OCL4DNN_CONFIG_PATH parameter.
OpenCL program build log: dnn/dummy
Status -11: CL_BUILD_PROGRAM_FAILURE
-cl-no-subgroup-ifp
Error in processing command line: Don't understand command line argument "-cl-no-subgroup-ifp"!
(1, 44, 46, 46)
[INFO]Model predicts time: 15.411957502365112
Press any key to continue . . .
运行也有结果,这个error是什么?
现在只有Ubuntu系统了,费劲心思配置成openpose,显存不够额,emmm...再试试cpu版的吧 唉
Ubuntu20 Alphapose
安装Anaconda
按照github上的步骤(传送门),注意依赖、gcc<6.0等,笔者选择conda方式安装,按照说明下载训练好的模型等,测试成功
**@**:~/***/AlphaPose$ conda activate alphapose
(alphapose) **@***:~/***/AlphaPose$ python scripts/demo_inference.py --cfg configs/coco/resnet/256x192_res50_lr1e-3_1x.yaml --checkpoint pretrained_models/fast_res50_256x192.pth --indir examples/demo/ --outdir examples/results/ --save_img --format cmu --sp
Loading YOLO model..
Loading pose model from pretrained_models/fast_res50_256x192.pth...
100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 4/4 [00:01<00:00, 3.23it/s]
===========================> Finish Model Running.
===========================> Rendering remaining images in the queue...
===========================> If this step takes too long, you can enable the --vis_fast flag to use fast rendering (real-time).
===========================> Rendering remaining 0 images in the queue...
Results have been written to json.
(alphapose) **@***:~/***/AlphaPose$