Train/Test set split

[Spritea]

Hi, could you please explain how the Train/Test set is splitted?

The paper mentions that 18 videos are used as training set and 3 as test set. But it looks like code of this repo does not specify the train/test split.

The default argument of data_txt_path_train in main.py is ./datasets/data_path/refer-kitti.train, but there is no infomation about the folder/file refer-kitti.train.

[wudongming97]

For example, the folder/file format 'KITTI/training/image_02/0001/000000.png' in refer-kitti.train follows the default setting of official KITTI.

[Spritea]

Thanks for reply! Could you specify which sequences are in the train set, and which sequences are in the test set?

[wudongming97]

You can check ./datasets/data_path, where refer-kitti.train has training sequences, and seqmap.txt has testing sequences.