End-to-end Holistic 3D Scene Understanding with Attention

Mert Kiray*, Halil Eralp Kocas*, Yinyu Nie

Introduction

This repo contains training, testing, evaluation, and visualization code of our project End-to-end Holistic 3D Scene Understanding with Attention. Our proposed architectures are GAT, Decoder only, and Complete Transformer to replace SGCN from Holistic3D.

If you encounter any issues running this repo, please email:

[email protected] or [email protected]

CUDA Adaptation

Please make sure you adapt the compatible gencode for CUDA_NVCC_FLAGS under external/mesh_fusion/libfusiongpu/CMakeLists.txt and external/ldif/ldif2mesh/build.sh according to your CUDA version and GPU architecture.

Install

Please note that the following Python, gcc, and CUDA versions are used to compile and run this project:

Python=3.7 & gcc=8 & g++=8 & CUDA 10.2

Please make sure to install CUDA NVCC on your system first. then run the following:

sudo apt install xvfb ninja-build freeglut3-dev libglew-dev meshlab
pip install -r requirements.txt
python project.py build

When running python project.py build, the script will run external/build_gaps.sh which requires password for sudo privilege for apt-get install. Please make sure you are running with a user with sudo privilege. If not, please reach your administrator for installation of these libraries and comment out the corresponding lines then run python project.py build.

Data preparation

We follow Holistic3D to use SUN-RGBD to train our GAT, Decoder only Transformer, Complete Transformer architectures and use Pix3D to train Holistic3D's Local Implicit Embedding Network (LIEN) with Local Deep Implicit Functions (LDIF) decoder.

Preprocess SUN-RGBD data

Please follow Total3DUnderstanding to directly download the processed train/test data.

In case you prefer processing by yourself or want to evaluate 3D detection (To ultilize the evaluation code of Coop, Holistic3D modified the data processing code of Total3DUnderstanding to save parameters for transforming the coordinate system from Total3D back to Coop), please follow these steps:

Follow Total3DUnderstanding to download the raw data.
According to issue #6 of Total3DUnderstanding, there are a few typos in json files of SUNRGBD dataset, which is mostly solved by the json loader. However, one typo still needs to be fixed by hand. Please find {"name":""propulsion"tool"} in data/sunrgbd/Dataset/SUNRGBD/kv2/kinect2data/002922_2014-06-26_15-43-16_094959634447_rgbf000089-resize/annotation2Dfinal/index.json and remove ""propulsion.
Process the data by
```
python -m utils.generate_data
```

Preprocess Pix3D data

Please follow these steps to generate the train/test data:

Download the Pix3D dataset to data/pix3d/metadata
Run below to generate the train/test data into 'data/pix3d/ldif'
```
python utils/preprocess_pix3d4ldif.py
```

Training and Testing

We use wandb for logging and visualization. You can register a wandb account and login before training by wandb login. In case you don't need to visualize the training process, you can put WANDB_MODE=dryrun before the commands bellow.

We manage parameters of each experiment with configuration files (configs/****.yaml). We first follow Holistic3D to pretrain each individual module, then jointly finetune the full model with additional physical violation loss.

Pretraining of Initial Estimation Stage

We use the pretrained checkpoint of Total3DUnderstanding to load weights for ODN. Please download and rename the checkpoint to out/pretrained_models/total3d/model_best.pth. Other modules can be trained then tested with the following steps:

Train LEN by:
```
python main.py --config configs/layout_estimation.yaml
```
The pretrained checkpoint can be found at out/layout_estimation/[start_time]/model_best.pth
Train LIEN + LDIF by:
```
python main.py --config configs/ldif.yaml
```
The pretrained checkpoint can be found at out/ldif/[start_time]/model_best.pth (alternatively, you can download the pretrained model here, and unzip it into out/ldif/20101613380518/)

The training process is followed with a quick test without ICP and Chamfer distance evaluated. In case you want to align mesh and evaluate the Chamfer distance during testing:
```
python main.py --config configs/ldif.yaml --mode train
```
The generated object meshes can be found at out/ldif/[start_time]/visualization

Training of Holistic3D

Replace the checkpoint directories of LEN and LIEN in configs/total3d_ldif_gcnn.yaml with the checkpoints trained above, then train SGCN by:
```
python main.py --config configs/total3d_ldif_gcnn.yaml
```
The pretrained checkpoint can be found at out/total3d/[start_time]/model_best.pth
Replace the checkpoint directory in configs/total3d_ldif_gcnn_joint.yaml with the one trained in the last step above, then train the full model by:
```
python main.py --config configs/total3d_ldif_gcnn_joint.yaml
```
The trained model can be found at out/total3d/[start_time]/model_best.pth
The training process is followed with a quick test without scene mesh generated. In case you want to generate the scene mesh during testing (which will cost a day on 1080ti due to the unoptimized interface of LDIF CUDA kernel):
```
python main.py --config configs/total3d_ldif_gcnn_joint.yaml --mode train
```
The testing results can be found at out/total3d/[start_time]/visualization

Training of Graph Attention Network

Replace the checkpoint directories of LEN and LIEN in configs/att_total3d_ldif_gcnn.yaml with the checkpoints trained above, then train SGCN by:
```
python main.py --config configs/att_total3d_ldif_gcnn.yaml
```
The pretrained checkpoint can be found at out/total3d/att/[start_time]/model_best.pth
Replace the checkpoint directory in configs/att_total3d_ldif_gcnn_joint.yaml with the one trained in the last step above, then train the full model by:
```
python main.py --config configs/att_total3d_ldif_gcnn_joint.yaml
```
The trained model can be found at out/total3d/att/[start_time]/model_best.pth
The training process is followed with a quick test without scene mesh generated. In case you want to generate the scene mesh during testing (which will cost a day on 1080ti due to the unoptimized interface of LDIF CUDA kernel):
```
python main.py --config configs/att_total3d_ldif_gcnn_joint.yaml --mode train
```
The testing results can be found at out/total3d/att/[start_time]/visualization

Training of Decoder only Transformer

Replace the checkpoint directories of LEN and LIEN in configs/transformer_total3d.yaml with the checkpoints trained above, then train SGCN by:
```
python main.py --config configs/transformer_total3d.yaml
```
The pretrained checkpoint can be found at out/total3d/trans/[start_time]/model_best.pth
Replace the checkpoint directory in configs/transformer_total3d_joint.yaml with the one trained in the last step above, then train the full model by:
```
python main.py --config configs/transformer_total3d_joint.yaml
```
The trained model can be found at out/total3d/trans/[start_time]/model_best.pth
The training process is followed with a quick test without scene mesh generated. In case you want to generate the scene mesh during testing (which will cost a day on 1080ti due to the unoptimized interface of LDIF CUDA kernel):
```
python main.py --config configs/transformer_total3d_joint.yaml --mode train
```
The testing results can be found at out/total3d/trans/[start_time]/visualization

Training of Complete Transformer Network

Replace the checkpoint directories of LEN and LIEN in configs/transformer_enc_dec_total3d.yaml with the checkpoints trained above, then train SGCN by:
```
python main.py --config configs/transformer_enc_dec_total3d.yaml
```
The pretrained checkpoint can be found at out/total3d/transformer_enc_dec/[start_time]/model_best.pth
Replace the checkpoint directory in configs/transformer_enc_dec_total3d_joint.yaml with the one trained in the last step above, then train the full model by:
```
python main.py --config configs/transformer_enc_dec_total3d_joint.yaml
```
The trained model can be found at out/total3d/transformer_enc_dec/[start_time]/model_best.pth
The training process is followed with a quick test without scene mesh generated. In case you want to generate the scene mesh during testing (which will cost a day on 1080ti due to the unoptimized interface of LDIF CUDA kernel):
```
python main.py --config configs/transformer_enc_dec_total3d_joint.yaml --mode train
```
The testing results can be found at out/total3d/transformer_enc_dec/[start_time]/visualization

Testing

The training process above already include a testing process. In case you want to test LIEN+LDIF or full model by yourself:
- For LDIF:
```
 python main.py --config out/ldif/[start_time]/out_config.yaml --mode test
```
- For Holistic3D:
```
 python main.py --config out/total3d/[start_time]/out_config.yaml --mode test
```
  The results will be saved to out/total3d/[start_time]/visualization and the evaluation metrics will be logged to wandb as run summary.
- For GAT / Decoder Only Transformer / Complete Transformer architectures:
```
python main.py --config out/total3d/[att/trans/transformer_enc_dec]/[start_time]/out_config.yaml --mode test
```
  The results will be saved to out/total3d/[att/trans/transformer_enc_dec]/[start_time]/visualization and the evaluation metrics will be logged to wandb as run summary.

Evaluate 3D object detection with modified matlab script from Coop:

external/cooperative_scene_parsing/evaluation/detections/script_eval_detection.m

Before running the script, please specify the following parameters:

SUNRGBD_path = 'path/to/SUNRGBD';
result_path = 'path/to/experiment/results/visualization';

Visualize the i-th 3D scene interacively by

For Holistic3D:

python utils/visualize.py --result_path out/total3d/[start_time]/visualization --sequence_id [i]

or save the 3D detection result and rendered scene mesh by

python utils/visualize.py --result_path out/total3d/[start_time]/visualization --sequence_id [i] --save_path []

In case you do not have a screen:

python utils/visualize.py --result_path out/total3d/[start_time]/visualization --sequence_id [i] --save_path [] --offscreen

For GAT / Decoder only / Complete Transformer architectures:

 python utils/visualize.py --result_path out/total3d/[att/trans/transformer_enc_dec]/[start_time]/visualization --sequence_id [i]

or save the 3D detection result and rendered scene mesh by

python utils/visualize.py --result_path out/total3d/[att/trans/transformer_enc_dec]/[start_time]/visualization --sequence_id [i] --save_path []

In case you do not have a screen:

python utils/visualize.py --result_path out/total3d/[att/trans/transformer_enc_dec]/[start_time]/visualization --sequence_id [i] --save_path [] --offscreen

Visualize the detection results from a third person view with modified matlab script from Coop:

external/cooperative_scene_parsing/evaluation/vis/show_result.m

Before running the script, please specify the following parameters:

SUNRGBD_path = 'path/to/SUNRGBD';
save_root = 'path/to/save/the/detection/results';
paths = {
    {'path/to/save/detection/results', 'path/to/experiment/results/visualization'}, ...
    {'path/to/save/gt/boundingbox/results'}
};
vis_pc = false; % or true, if you want to show cloud point ground truth
views3d = {'oblique', 'top'}; % choose prefered view
dosave = true; % or false, please place breakpoints to interactively view the results.

Citation

We thank the following great works:

Holistic3D for their SGCN, LIEN and LDIF.
Total3DUnderstanding for their well-structured code. We construct our network based on their well-structured code.
Coop for their dataset. We used their processed dataset with 2D detector prediction.
LDIF for their novel representation method. We ported their LDIF decoder from Tensorflow to PyTorch.
Graph R-CNN for their scene graph design. We adopted their GCN implemention to construct our SGCN.
Occupancy Networks for their modified version of mesh-fusion pipeline.

If you find them helpful, please cite:

@article{zhang2021holistic,
  title={Holistic 3D Scene Understanding from a Single Image with Implicit Representation},
  author={Zhang, Cheng and Cui, Zhaopeng and Zhang, Yinda and Zeng, Bing and Pollefeys, Marc and Liu, Shuaicheng},
  journal={arXiv preprint arXiv:2103.06422},
  year={2021}
}
@InProceedings{Nie_2020_CVPR,
author = {Nie, Yinyu and Han, Xiaoguang and Guo, Shihui and Zheng, Yujian and Chang, Jian and Zhang, Jian Jun},
title = {Total3DUnderstanding: Joint Layout, Object Pose and Mesh Reconstruction for Indoor Scenes From a Single Image},
booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2020}
}
@inproceedings{huang2018cooperative,
  title={Cooperative Holistic Scene Understanding: Unifying 3D Object, Layout, and Camera Pose Estimation},
  author={Huang, Siyuan and Qi, Siyuan and Xiao, Yinxue and Zhu, Yixin and Wu, Ying Nian and Zhu, Song-Chun},
  booktitle={Advances in Neural Information Processing Systems},
  pages={206--217},
  year={2018}
}	
@inproceedings{genova2020local,
    title={Local Deep Implicit Functions for 3D Shape},
    author={Genova, Kyle and Cole, Forrester and Sud, Avneesh and Sarna, Aaron and Funkhouser, Thomas},
    booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
    pages={4857--4866},
    year={2020}
}
@inproceedings{yang2018graph,
    title={Graph r-cnn for scene graph generation},
    author={Yang, Jianwei and Lu, Jiasen and Lee, Stefan and Batra, Dhruv and Parikh, Devi},
    booktitle={Proceedings of the European Conference on Computer Vision (ECCV)},
    pages={670--685},
    year={2018}
}
@inproceedings{mescheder2019occupancy,
  title={Occupancy networks: Learning 3d reconstruction in function space},
  author={Mescheder, Lars and Oechsle, Michael and Niemeyer, Michael and Nowozin, Sebastian and Geiger, Andreas},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
  pages={4460--4470},
  year={2019}
}

Name		Name	Last commit message	Last commit date
Latest commit History 46 Commits
configs		configs
data		data
demo/inputs		demo/inputs
external		external
figures		figures
libs		libs
models		models
net_utils		net_utils
utils		utils
.gitignore		.gitignore
End_to_End_Holistic_3D_Scene_Understanding_with_Attention_from_a_Single_Image.pdf		End_to_End_Holistic_3D_Scene_Understanding_with_Attention_from_a_Single_Image.pdf
LICENSE		LICENSE
README.md		README.md
demo.py		demo.py
main.py		main.py
project.py		project.py
requirements.txt		requirements.txt
test.py		test.py
test_epoch.py		test_epoch.py
train.py		train.py
train_epoch.py		train_epoch.py

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

End-to-end Holistic 3D Scene Understanding with Attention

Introduction

CUDA Adaptation

Install

Data preparation

Preprocess SUN-RGBD data

Preprocess Pix3D data

Training and Testing

Pretraining of Initial Estimation Stage

Training of Holistic3D

Training of Graph Attention Network

Training of Decoder only Transformer

Training of Complete Transformer Network

Testing

Citation

About

Releases

Packages

Languages

License

mertkiray/EndtoEnd3DSceneUnderstandingWithAttention

Folders and files

Latest commit

History

Repository files navigation

End-to-end Holistic 3D Scene Understanding with Attention

Introduction

CUDA Adaptation

Install

Data preparation

Preprocess SUN-RGBD data

Preprocess Pix3D data

Training and Testing

Pretraining of Initial Estimation Stage

Training of Holistic3D

Training of Graph Attention Network

Training of Decoder only Transformer

Training of Complete Transformer Network

Testing

Citation

About

Resources

License

Stars

Watchers

Forks

Releases

Packages 0

Languages

Packages