search

haofeixu / aanet

[CVPR'20] AANet: Adaptive Aggregation Network for Efficient Stereo Matching
Apache License 2.0
524 stars 102 forks source link

performance without pseudo-labels #7

Closed waychin-weiqin closed 4 years ago

waychin-weiqin commented 4 years ago

Hi,

Great work and thanks for sharing the code. May I ask if you have the numbers for the performance of your model when trained with only the sparse KITTI ground truth or in other words without the pseudo-labels provided by other network?

waychin-weiqin commented 4 years ago

Don't worry about it. I found it in the paper.

waychin-weiqin commented 4 years ago

Hi,

May I know if the results of your work without pseudo ground truth you shared in your paper is evaluated by uploading the results to KITTI benchmark or by using the matlab script provided by KITTI?

If you are evaluating using matlab, can you provide the list of testing images? Thank you

haofeixu commented 4 years ago

Hi @SamChuah ,

The results are evaluated by our python code https://github.com/haofeixu/aanet/blob/5e2359cfea9c38ebcddf5e750ec9ad1ac92725fc/model.py#L258-L262. The validation split is available at https://github.com/haofeixu/aanet/blob/master/filenames/KITTI_2015_val.txt

waychin-weiqin commented 4 years ago

Thanks for the swift reply. Do you know if there is any difference between your python code and the matlab script provided by KITTI?

haofeixu commented 4 years ago

The results should be very close if not exactly same, I cannot recall precisely. You can double check on your side by comparing the evaluation results.

waychin-weiqin commented 4 years ago

Thank you very much :)