/3D-Object-Reconstruction-from-Multi-View-Monocular-RGB-images

Hybrid Ensemble Approach For 3D Object Reconstruction from Multi-View Monocular RGB images

Primary LanguagePythonMIT LicenseMIT

3D-Model-Reconstruction from 2D images

A Novel Hybrid Ensemble Approach For 3D Object
Reconstruction from Multi-View Monocular RGB images for Robotic Simulations

Ajith Balakrishnan, Dr. Sreeja S

(1) Architecture

Arch Image

(2) STSO-JTSO Algorithm

Teaser Image

(3) Sample Results

Teaser Image

(4) Data

3D-R2N2 Dataset

https://github.com/chrischoy/3D-R2N2

LSM Dataset

https://github.com/akar43/lsm

(5) Released Model

Trained on 3D-R2N2 dataset

will release soon

(6) Requirements

python 3.5

tensorflow 1.11.0

numpy 1.13.3

scipy 0.19.0

matplotlib 2.0.2

skimage 0.13.0

(7) Run

Training

python main_AttSets.py

Test Demo (Download released model first)

python demo_AttSets.py

(8) Citation

If you use the paper or code for your research, please cite:

(9) Special Thanks To

https://github.com/Yang7879/AttSets // https://github.com/chrischoy/3D-R2N2 // https://arxiv.org/pdf/1901.11153.pdf