bubbliiiing / Semantic-Segmentation

I will upload many semantic segmentation models to this repository for you to learn
MIT License
438 stars 157 forks source link

segnet 换成自己的影像出现错误 #1

Open ttw108 opened 4 years ago

ttw108 commented 4 years ago

您好, 参照您的segnet, 斑马线上没问题,可以有效训练+预测,甚至网上抓的图片也能成功预测。 因此,用自身影像以labelme label后(0:bk 1:target) , 将jpg, png置于./dataset2/jpg 及./dataset2/png 后,开始run train.py... 出现error https://prnt.sc/r57q5l File "", line 27, in generate_arrays_from_file IndexError: too many indices for array

相关档案于此,可否请您协助剖析。万分感谢! https://drive.google.com/file/d/1RlgNxoIwhJj0liYarbQVZEn8397tQ_PL/view?usp=sharing

ttw108 commented 4 years ago

已解決, labelme_json_to_dataset 後的PNG 都均要将 label 转成全局的 类别才行。

clearliu777 commented 3 years ago

请问这个全局的label是怎么转变的

bubbliiiing commented 3 years ago

按视频