nice-mee
commented
2 months ago The yolov8 demo in this project is basically a simplified version of official yolov8 demo. Theoretically it should work fine with official yolov8 onnx file, provided that you have correct label file in
But this project is only intended for RV1103/1106, I deleted many lines unrelated to RV1103/1106. You will definitely run into issues when trying to deploy it onto any platform other than RV1103/1106.
In the yolov8_pose demo I used a custom onnx model converted from a modified script, something similar to this, except that I also modified the pose detection branch. Unfortunately, I cannot find the Python scripts anymore.
这个项目中的 yolov8 演示基本上是 官方 yolov8 演示 的简化版本。理论上，只要你有正确的标签文件，它应该可以与官方的 yolov8 onnx 文件正常工作。
但是，这个项目仅适用于 RV1103/1106，我删除了许多与 RV1103/1106 无关的代码。如果你尝试将其部署到 RV1103/1106 以外的任何平台上，你肯定会遇到问题。
在 yolov8_pose 演示中，我使用了一个从修改后的脚本转换来的 onnx 模型，类似于 这个官方仓库，只是我还修改了姿态检测分支。不幸的是，我已经找不到相关的 Python 脚本了。
请问模型转换有什么技巧？用官方的yolov8不可用！