Chilicyy

以下是我们做的相关实验结果，从结果上看训练时长差距不大，评估时长由于采用不同的评估方式，这块后续会进行优化。 框架 | 输入 | 数据集 | 训练集/验证集数量 | 资源 | 训练时间(1e) | 评估时间 -- | -- | -- | -- | -- | -- | -- yolov5n | 640*640...

@JiaPai12138 您好，如果可以的话，麻烦将workers数设置更大一些，确认下是否数据加载速度较慢的原因呢

@yuvraajj Hi, we have fixed the bug when resuming training. You could train the model again.

Hi @buyukkanber , `do_coco_metric ` means using pycocotools to evaluate metrics, which is common applied in papers and competitions, while `do_pr_metric ` means using evaluation scripts similar to yolov5 and...

_apply(self, fn)是PyTorch中nn.Module的一个内部方法，它用于递归地对模块的所有子模块应用一个函数fn。这个函数通常用于模块的初始化或者将模块的所有参数转移到GPU或CPU。 在你给出的代码中，_apply(self, fn)被重写以适应特定的模块。除了默认的对所有子模块应用fn之外，它还将fn应用于self.detect.stride和self.detect.grid。这可能是因为这两个属性在模块的计算中起着关键的作用，因此当模块的其他部分发生变化时（例如，当模块被移动到GPU时），这两个属性也需要相应地更新。 总的来说，_apply(self, fn)的作用是将一个函数应用于模块的所有子模块，以及在这个特定模块中，还应用于self.detect.stride和self.detect.grid。

Thanks for your attention. We'll consider it and add this feature.

Hi, it seems that you are using Pillow >= 10.0.0 , you can try the following method. `pip uninstall Pillow` `pip install Pillow==9.5.0`

Hi @sisgrad , we do experiments by using only `--distill`, and using coco_checkpoint for pretrained model. The performance of self-distillation may be vary from different tasks, and you could try...

Hi, P5 models support fuse_ab training, but P6 models do not support fuse_ab training currently. If you want to train P5 models like yolov6n/s/m/l, you can add anchor_init refer to...

P6 models add an extra stage on the top of the backbone to have a feature (C6) at a higher level for detecting extra- large objects.