Chilicyy comments

Results 107 comments of


                                            Chilicyy

Model training and evaluation problems

Thanks for your suggestion! We'll mark it down and try to add this feature.

yolov6 nano在训练的最后100个epoch有特殊操作吗?

您好，在最后100个epoch没有其他特殊操作的。麻烦提供下这边训练的环境以及相关的参数，包括epoch数、训练卡数以及batchsize等。

yolov6 nano在训练的最后100个epoch有特殊操作吗?

> > 您好，在最后100个epoch没有其他特殊操作的。麻烦提供下这边训练的环境以及相关的参数，包括epoch数、训练卡数以及batchsize等。 > > 請問所有的結果都是在混合精度訓練開啓的情況下得到的嗎？另外還有一個問題，爲什麼dataset讀取coco數據時不使用pycocotools，而是手動讀取了全部數據，相比於pycocotools多了1000多張圖像，這樣對精度有影響嗎？期待您的解答您好，感谢您的关注。我们的训练过程默认是开启混合精度训练的。关于数据读取的问题，我们手动读取是为了兼容用户自定义的数据输入格式，方便代码维护，还没尝试过使用pycocotools读取数据，不确定对精度是否有影响。如果您这边有相关的实验结果，也可以告知我们，谢谢！

ModuleNotFoundError: No module named 'tools.eval'

Thanks for your attention! You can try adding the root directory of the project into the PYTHONPATH, like export PYTHONPATH=$PYTHONPATH:/xxx/YOLOv6

Module not found

Thanks for your attention! You can try to add the directory of YOLOv6 to $PYTHONPATH.

about YOLOv6L

@BowieHsu Thanks for your attention! We 'll release large size models for YOLOv6, and the results shown in the table seem much different from ours.

Is there a way to change model_fitness similar to YOLOv5 so you can increase recall or precision?

@sarmientoj24 Thanks for your attention! Currently we don't support this feature. We will consider your suggestion and add it in subsequent releases.

Feature: Added camera support for realtime Object Detection

Thanks for your attention! It'll be appreciated that you can add this feature based on our latest code, because we have updated some relevant scripts to fit different data formats.

AssertionError: ../coco/images/train2017 is an invalid directory path!

Maybe you can try changing the relative path to absolute path in data/coco.yaml.

Adding U-Net backbone before EfficientRep

Thx for your attention! If you want to add an extra network before the backbone, you need to create a pyfile for your own network structure , and then put...