-
Notifications
You must be signed in to change notification settings - Fork 135
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Fixcommit #25 #49
base: main
Are you sure you want to change the base?
Fixcommit #25 #49
Conversation
…-epoch.pth Rename to prevent filename error on Windows platform.
…e on Win platform,numpy readonly
…e on Win platform,numpy readonly
请问python版本是3.8吗 |
是的,是python3.8.19 |
具体可以看我的conda环境版本: |
已经完成了onnx在Windows和Linux上的训练与测试,同时也完成了ncnn在Linux上可执行文件的编译推理,均正常 本来我只是实验室三次考核结合的FastestDet,看到很多坑就修复了一下 |
好的谢谢. |
后面我换回官方的用3.8安装了
|
有一些包版本仅限某些python的,onnxsimple这个在FastestDet确实得低版本,没办法,FastestDet就是这个时候写的 |
随着多个依赖的更新修复多处兼容性、多平台问题:
1、pip 不再拥有原requirements.txt内的onnxruntime==1.11.1,修改为1.16.0,
更高的版本在Windows上onnxruntime官方已经确认有问题:[1.16.2] ONNX not built correctly for Windows
2、在train.py修改保存模型名以修复Windows平台不支持冒号作为文件名的问题,修复后可以close的关联Issue有:
AssertionError: 请指定正确的配置文件路径
README weight_AP05:0.253207_280-epoch.pth Colon Issues
train.py导出路径
pth文件命名带了冒号
请问大佬这训练过程是否有出错的地方,一直找不到训练出来的模型
无法保存模型权重
3、onnx示例添加防止传入后预测框丢失的情况,不进行判断没有目标时进行nms会报numpy数组维度错误:
IndexError: too many indices for array: array is 1-dimensional, but 2 were indexed.
4、onnx示例添加防止推理要绘制的内容带numpy的只读属性的图像矩阵复制,会报不可修改:
img marked as output argument, but provided NumPy array marked as readonly
opencv官方库里cv::rectangle、cv::circle(c++的,对应到Python里就是cv2.rectangle和cv2.circle)没修复这个问题的Issue:
Document OpenCV 4.9 Python changes behavior with readonly numpy arrays