search

FangGet / PackNet-SFM-PyTorch

A Pytorch implementation(unofficial) for paper "PackNet-SfM: 3D Packing for Self-Supervised Monocular Depth Estimation"
GNU General Public License v3.0
21 stars 7 forks source link

The model can be adapted accordingly to your current version as the authors have released the code. #7

Closed rvarun7777 closed 4 years ago

rvarun7777 commented 4 years ago

https://github.com/TRI-ML/packnet-sfm

FangGet commented 4 years ago

thanks for reminding, I'll do it asap.

vicentowang commented 4 years ago

I compared the original with this , so many mistakes in this code. you may need clearly read the paper,i recommend.

rvarun7777 commented 4 years ago

@vicentowang I mean, first off it's a good job and a very good try from @FangGet to reproduce the paper. Sometimes, it's not easy to implement architecture by just reading the paper. I hope the changes will be incorporated by @FangGet with the inference code. In our work FisheyeDistancenet we were able to achieve superior results with resnet itself. The gain from pack net is minimal though.

vicentowang commented 4 years ago

you are right ,but i insist that we should build it then rebuild. by the way, i read the paper only ,and coding the right architecture network. i just read the paper carefully . nothing big difference.

FangGet commented 4 years ago

thanks @rvarun7777 . thanks for your advice. @vicentowang I'll take some time to rebuild this project.