GDbbq

Results 23 comments of GDbbq

![image](https://github.com/PaddlePaddle/Paddle3D/assets/86715812/5688c2c5-93bf-465f-8371-7373106eb5f4) ![image](https://github.com/PaddlePaddle/Paddle3D/assets/86715812/14d93489-e9f9-4868-8fb2-0bd59a592f75) ![image](https://github.com/PaddlePaddle/Paddle3D/assets/86715812/4382388f-778a-45c9-8b95-376ebdb14e79)

你是用paddle提供onnx模型吗?建议检查下环境

想问下,楼主有测试过其他模型吗?

I had the same problem. I'm guessing it's something at the code level. The yolov5 model quantified by trtexec pqt can run normally, but the accuracy loss is serious. In...

nvidia device: 1080ti ![image](https://github.com/marcoslucianops/DeepStream-Yolo/assets/86715812/21d9c4da-b15b-4ec9-9419-bbaf73f4fb01)

我用gpu也没跑起来,这个模型似乎很消耗显存,24G都不够。。你可以换个模型试试

请问问题解决了吗,我也遇到类似的问题。 ![image](https://github.com/EasyDarwin/EasyPusher/assets/86715812/920eb9e6-30bb-42b9-be6b-4e9ac8b8172b)

已解决,可能是标签的问题,改用voc便签标注的方法即可

上述问题,已找到原因,就是自己的数据没做归一化

我使用ffmpeg-mpp 做硬解码, cpu利用率差不多,但是再用rga做缩放,cpu利用率反而会更高