-
当网络量化情况不好，并且推理框架允许的情况下，我们可以送一些算子回到fp32，这是量化部署的正确方法。
但是在最新的提交中，似乎代码中衡量量化敏感性的方法是直接在验证集上计算map，我想在目前的yolo6中，有将近100个层需要分析量化敏感性，这意味着整个验证过程需要持续近100次。我想这并不是一时半会能算完的。
一些更加高效的做法是，应当衡量的是网络中最后几个特征图的信号误差情况，而非网…
-
# 【PaddlePaddle Hackathon 4】模型套件开源贡献任务合集
（此 ISSUE 为 PaddlePaddle Hackathon 第四期活动的任务 ISSUE，更多详见 [【PaddlePaddle Hackathon 第四期】任务总览](https://github.com/PaddlePaddle/Paddle/issues/50629)）
注：开发请参考 [贡…
-
Any plans for pip install yolov6? I want to test 3rd party applications on Yolov6 model. Popular repos require pip for installation.
-
And is it possible to change it?
-
Thanks for your great work, is there a plan to support keypoint detection or just support quadrilateral box?
-
I'm using the DTLD traffic light dataset. The objects is normaly small.
When I use YOLOX-s, the mAP@0.5 achieve at 0.569. Yet the yolov6-s only achieve at 0.537.(0.475 if no loading pretrained pt.…
-
There is no except clause which is causing syntax error in training for try statement in line 105
-
Hi.
I followed the [Tutorials](https://github.com/meituan/YOLOv6/blob/main/docs/Train_custom_data.md) and attempted to train on a custom dataset.
However, due to a bug in the get_imgs_labels metho…
-
tensorrt two ways：
1、as the https://github.com/wang-xinyu/tensorrtx/tree/master/yolov5
pt->wts->tensorrt
i try the pt convert to wts use the gen_wts.py，but false, with error: AttributeErr…
-
AssertionError: Results do not correspond to current coco set