Skip to content

Code release for "3D-RelNet: Joint Object and Relation Network for 3D prediction"

Notifications You must be signed in to change notification settings

nileshkulkarni/relative3d

Repository files navigation

3D-RelNet: Joint Object and Relation Network for 3D prediction

Nilesh Kulkarni, Ishan Misra, Shubham Tulsiani, Abhinav Gupta.

Project Page

Teaser Image

Demo and Pre-trained Models

Please check out the interactive notebook suncg, interactive notebook nyu which shows reconstructions using the learned models. To run this, you'll first need to follow the installation instructions to download trained models and some pre-requisites.

Training and Evaluating

To train or evaluate the (trained/downloaded) models, it is first required to download the SUNCG dataset and preprocess the data and download the splits here. Please see the detailed README files for Training or Evaluation of models for subsequent instructions. Please note that these splits are different than the splits used by Factored3d

To train or evaluate on the NYUv2 dataset the (trained/downloaded) models, it is first required to download the NYU dataset and preprocess the data and download the splits here. Please see the detailed README files for Training or Evaluation of models for subsequent instructions.

Citation

If you use this code for your research, please consider citing:


@article{kulkarni20193d,
  title={3D-RelNet: Joint Object and Relational Network for 3D Prediction},
  author={Kulkarni, Nilesh
  and Misra, Ishan 
  and Tulsiani, Shubham
  and Gupta, Abhinav},
  journal={International Conference on Computer Vision (ICCV)}
  year={2019}
}