Issues with training and testing on the MSEC dataset

[id4su]

Hello, when I train on the dataset you proposed, everything goes relatively well. However, I have the following problem when I train on the MSEC dataset, can you help me? Thanks!

I'm sure my path settings are correct.

Actually, I found that in the training set, validation set and testing set of MSEC, five input maps correspond to one ground truth image, how can I solve this problem so that I can train and test normally.

[onpix]

Please try this setting:

# @package ds
class: img_dataset
name: MSEC
input:
  - /public/hywang26/ds/msec/train/INPUT_IMAGES/*_0.JPG
  - /public/hywang26/ds/msec/train/INPUT_IMAGES/*_N1.JPG
  - /public/hywang26/ds/msec/train/INPUT_IMAGES/*_N1.5.JPG
  - /public/hywang26/ds/msec/train/INPUT_IMAGES/*_P1.JPG
  - /public/hywang26/ds/msec/train/INPUT_IMAGES/*_P1.5.JPG
GT:
  - /public/hywang26/ds/msec/train/GT_IMAGES/*
  - /public/hywang26/ds/msec/train/GT_IMAGES/*
  - /public/hywang26/ds/msec/train/GT_IMAGES/*
  - /public/hywang26/ds/msec/train/GT_IMAGES/*
  - /public/hywang26/ds/msec/train/GT_IMAGES/*

which will extend GT 5 times, one for each input image subset.