LMM

@xiexinch Thanks a lot, but at the same time can the config file adapt to different input sizes, eg 192x192 network input.

Hi @xiexinch I'm very sorry, I don't need to answer the above questions. I have found the reason and solved it. It is the problem of model weight preservation. Can...

HI@xiexinch, thank you very much! Can I understand that it is only necessary to modify the output structure of the mobnetv3 network under the bankbon file and the input structure...

Hi, hello, I also encountered the same problem when running in unbantu18.04. If it is solved, can you provide a solution? Thank you!

好的，谢谢

好的，谢谢

嗯好的，谢谢！ @juncaipeng,但是我还想问一个问题，在我将数据集改了VOC_seg数据时，同时模型预处理输入改为了320*320到网络模型中，那么最终这个两个是否要修改对应上 backbone_indices: [2, 3] pp_out_channels: 1024

@MeowZheng Hi, can I understand that in the class_weigt weight loss, the weight with more pixels should appropriately reduce the weight coefficient, and if there are fewer pixels, increase the...

嗯好的，谢谢可以了

@jiangjiajun ，用于模型的推理部署