Sunflower7788 comments

Results 107 comments of


                                            Sunflower7788

TFT方法有办法修改损失函数为MAE、MSE之类的其他函数么

您好，由于问题超过一周，将自动关闭。若有问题可以重新打开，或者重新提issue。感谢理解。

安装Paddlex 2.0失败

欢迎尝试使用PaddleX 新版本进行训练，https://aistudio.baidu.com/intro/paddlex

是否考虑在EIseg中添加对Segment Anything模型的支持🤔

请关注即将发布的PaddleSeg v2.8新版本，将支持SAM。后续结合交互分割等工作，敬请期待。

运行upernet_vit报错了

请补充下运行命令

使用 modnet-mobilenetv2 模型，图像分割效果远离预期

请给出具体执行命令环境等

loss_ = -1 * np.power(1 - pro_, self._gamma) * np.log(pro_) is wrong

Yes,Iknow. In focal loss.py loss_ = -1 * np.power(1 - pro_, self.gamma) * np.log(pro) .I know the loss metric just a display and dont affact the model. But I think...

loss_ = -1 * np.power(1 - pro_, self._gamma) * np.log(pro_) is wrong

I think softmaxloss = -sum(yilog(pi)). But label is one hot. So softmaxloss = -log(pt)

PaddleOCR slim export model

您好，问题超过一周未回复将关闭。如有问题可重开issue.

The division of coco data is different from semi-supervised detection methods on COCO?

Thanks. But i want to know, is this the data division firstly proposed in this paper? Why this data division diffierent from semi-supervised detection for comparing the results of Boxes...

The division of coco data is different from semi-supervised detection methods on COCO?

> As I know, this dividion is firstly proposed in "Data Distillation: Towards Omni-Supervised Learning". Got it. Thanks.