wangguanzhong

Results 131 comments of wangguanzhong

把ppdet输出的results.json转成labelme可用的 json 格式

请向develop分支提交pr改动

PPTracking C++ not found bouding box when output video

我们在ubuntu 16，cuda10.2，cudnn7.6的环境上是ok的，看起来cmake命令中没有包含opencv，如果是ubuntu18及以上的版本，建议提前编译opencv然后再编译ppdet的c++代码

PPTracking C++ not found bouding box when output video

https://github.com/PaddlePaddle/PaddleDetection/tree/develop/deploy/pptracking/cpp 就是这份代码，近期没有更新过了

Hrnet pose just for fun？

部署再arm cpu上的话我们更推荐使用tinypose模型，https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/keypoint/tiny_pose#pipeline%E6%80%A7%E8%83%BD 这里有高通865上的测试数据 HRNet属于服务器端的模型，我们有测试T4上的性能可以参考https://github.com/PaddlePaddle/PaddleDetection/blob/develop/configs/keypoint/KeypointBenchmark.md#benchmark-on-server

关于PaddleDetection infer save result问题

https://github.com/PaddlePaddle/PaddleDetection/pull/6547 问题已修复 感谢反馈

fix json_results.py

Please check the tools/infer.py in picodet. There is an error reported in PaddleDetection-release-ci

/one_hot_kernel.cu:38 Assertion `p_in_data[idx] >= 0 && p_in_data[idx] < depth` failed. Illegal index value, Input(input) value should be greater than or equal to 0, and less than depth [2], but received [5],CUDA error(719), unspecified launch failure.

https://github.com/PaddlePaddle/PaddleDetection/blob/release/2.4/configs/datasets/coco_instance.yml#L2 需要检查下这里的配置是否和实际数据集一致

[PaddlePaddle Hackathon] Task 72: Add PAA

建议向develop分支提交pr，当前改动量太大了

模型评估类别显示不全

检查下验证集中是否5个类别都包含了

Restructuring training visualization and adding support for Weights & Biases Logger

Thx for your contribution. Could you add the function to PaddleMIX models https://github.com/PaddlePaddle/PaddleMIX/tree/develop