takaniwa / DSNet

a novel real-time model in semantic segmentation
MIT License
55 stars 0 forks source link

Can this model be applied to custom datasets? #5

Open XXXCARREY opened 2 days ago

XXXCARREY commented 2 days ago

For example,I have a picture and I want to get semantic masks of this picture. What should I do?

takaniwa commented 2 days ago

For example,I have a picture and I want to get semantic masks of this picture. What should I do? 注意到你是中国人,我就用中文回复你了,方便一点哈哈. 关于你的问题,如果直接用我给的的模型推理你的图片,可能不会有好的结果.建议你读取预训练模型之后进行微调,在你的数据集上以小学习率进行训练. 你需要在dataset里面新建一个能读取你数据集格式的dataset格式. 需要注意的是,这个代码目前好像有点问题,直接使用好像跑不起来(应该是因为创建项目的时候一个一个创建添加的),我需要修改一下训练脚本,由于最近实在是没有时间,请给我一些时间进行排查和修改,你可以先收藏以免到时候找不到,修改之后我也会给你留言的, 谢谢你的关注.

XXXCARREY commented 2 days ago

For example,I have a picture and I want to get semantic masks of this picture. What should I do? 注意到你是中国人,我就用中文回复你了,方便一点哈哈. 关于你的问题,如果直接用我给的的模型推理你的图片,可能不会有好的结果.建议你读取预训练模型之后进行微调,在你的数据集上以小学习率进行训练. 你需要在dataset里面新建一个能读取你数据集格式的dataset格式. 需要注意的是,这个代码目前好像有点问题,直接使用好像跑不起来(应该是因为创建项目的时候一个一个创建添加的),我需要修改一下训练脚本,由于最近实在是没有时间,请给我一些时间进行排查和修改,你可以先收藏以免到时候找不到,修改之后我也会给你留言的, 谢谢你的关注.

噢噢好的,感谢