search

CSAILVision / semantic-segmentation-pytorch

Pytorch implementation for Semantic Segmentation/Scene Parsing on MIT ADE20K dataset
http://sceneparsing.csail.mit.edu/
BSD 3-Clause "New" or "Revised" License
4.96k stars 1.1k forks source link

Error using custom data set #198

Closed jeewa985 closed 5 years ago

jeewa985 commented 5 years ago

I have encountered an error during evaluation for my own data set. anyone`s comments and suggestions on the following regard is highly appreciated. Segmenatation_6

jeewa985 commented 5 years ago

I have checked the shape of img, seg_color and pre_clorwhich are not same. seg_color and and pred_color only having one row it seems. training phase was successfully completed with about 100% accuracy. any commnets or suggestions on this is highly appreciated. Segmenatation_7

sainatarajan commented 5 years ago

@jeewa985 Hi, how did you prepare your custom dataset. What changes in the code you have to make? I would like to try on my own dataset. I have a directory of images and another directory of masks.

jeewa985 commented 5 years ago

@sainatarajan

custom dataset:

  1. create respective annotations images in gray scale including class indexes
  2. Make your own .odgt files for train and validation on your dataset.

Changes in code:

  1. Nothing much but you have to change .yaml file inside your config folder accordingly.

Hope this helps

simaiden commented 4 years ago

@sainatarajan

custom dataset:

1. create respective annotations images in gray scale including class indexes

2. Make your own .odgt files for train and validation on your dataset.

Changes in code:

1. Nothing much but you have to change .yaml file inside your config folder accordingly.

Hope this helps

Can you help me with the annotations images? I don't understand how can labelling the different classes using only one gray scale image, and also where i have to put the class indexes.

Thanks!