Classweighting of SegNet?

[qiaobaiyisheng]

If I use the SenNet, do I need to compute the classweighting mentioned in the paper?

[Qirui-Y]

Hello, are you solved problem ? Can you share your config file?

[qiaobaiyisheng]

model: arch: segnet data: dataset: camvid train_split: train val_split: val img_rows: 360 img_cols: 480 path: /your/path/pytorch-semseg-master/data/CamVid/ training: train_iters: 150000 batch_size: 8 val_interval: 100 n_workers: 16 print_interval: 50 optimizer: name: 'sgd' lr: 1.0e-10 weight_decay: 0.0005 momentum: 0.99 loss: name: 'cross_entropy' size_average: False lr_schedule: resume: segnet_camvid_model.pkl

I didn't found where to compute the classweighting in this code, so directly run the program with the config file above. The result on CamVid is about 0.58(mIOU), which is close to the experimental result in the proposed SegNet paper.

[Qirui-Y]

OK，thank you very much,I am try it!