search

haofeixu / aanet

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

Test results on Middlebury Stereo Evaluation - Version 3 #36

Closed baoguoma closed 4 years ago

baoguoma commented 4 years ago

Hi

I hope everything is OK. Recently I tested AANet on Middlebury Stereo Evaluation - Version 3, but I got very bad results. What could be the reason? I used every weight in pretrained model but none of them give the good results. Do I need to train AANet on Middlebury Stereo Evaluation - Version 3?

Thank you very much.

haofeixu commented 4 years ago

You may need to fine-tune our model on the target domain as the cross-dataset generalization ability may be not perfect.

baoguoma commented 4 years ago

You may need to fine-tune our model on the target domain as the cross-dataset generalization ability may be not perfect.

Thank you so much for your help, appreciate it!

mariocyrus commented 3 years ago

Hello! @haofeixu how would you addres the fine-tuning of your model? Could you please share some tips/steps?

Thank you :)