search

zzangjinsun / NLSPN_ECCV20

Park et al., Non-Local Spatial Propagation Network for Depth Completion, ECCV, 2020
MIT License
321 stars 57 forks source link

Request for released model #8

Closed youzunzhi closed 3 years ago

youzunzhi commented 3 years ago

Hi! Thank you for sharing your source code! I wonder if you can also share your trained models of both datasets which achieve the performance in your paper so I can directly run the testing code. Thank you very much!

zzangjinsun commented 3 years ago

Unfortunately, after code version up, the pre-trained weights are not directly compatible with the new codes. I will train full-model later because I have a limited number of GPUs now (used for other works now)

youzunzhi commented 3 years ago

That's cool. Thanks.

JUGGHM commented 3 years ago

Unfortunately, after code version up, the pre-trained weights are not directly compatible with the new codes. I will train full-model later because I have a limited number of GPUs now (used for other works now)

Seems it's a common confusing problem: code released should be neat, but not the same when good experiment results occur.

zzangjinsun commented 3 years ago

@youzunzhi If you don't mind to use a new checkpoint file trained with less number of batch size, I can share that file. It is trained with a small number of image batches for new code verification (slightly worse than the best result). Plz let me know your e-mail.

youzunzhi commented 3 years ago

@youzunzhi If you don't mind to use a new checkpoint file trained with less number of batch size, I can share that file. It is trained with a small number of image batches for new code verification (slightly worse than the best result). Plz let me know your e-mail.

Thank you so much! It's youzunzhi@gmail.com

zzangjinsun commented 3 years ago

Please check the e-mail. Thx!!

imransai commented 3 years ago

Hi Jinsun, Would you mind to send me the model trained with less batch sizes? Really appreciate it! My email id is saifimran06@gmail.com.

rpautrat commented 3 years ago

Hi Jinsun,

Thank you for the nice work and for sharing the code. I am also interested in your model trained with smaller batch sizes, as I am not sure to be able to wait that you release the bigger model. My email address is remi.pautrat@gmail.com. Thank you very much!!

zzangjinsun commented 3 years ago

I already released pre-trained models. Please refer to the https://github.com/zzangjinsun/NLSPN_ECCV20#pre-trained-models-and-results

rpautrat commented 3 years ago

Great, sorry for missing that! Thanks a lot for sharing them!