# 3D_Object_Reconstruction

**Repository Path**: zhangwx21/Object_Reconstruction

## Basic Information

- **Project Name**: 3D_Object_Reconstruction
- **Description**: A Novel Hybrid Ensemble Approach For 3D Object Reconstruction from Multi-View Monocular RGB images for Robotic Simulations
- **Primary Language**: Python
- **License**: Apache-2.0
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2021-09-06
- **Last Updated**: 2024-06-15

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# 3D_Object_Reconstruction

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

Code comes from: https://github.com/Ajithbalakrishnan/3D-Object-Reconstruction-from-Multi-View-Monocular-RGB-images

#### 算法架构
![Image text](https://gitee.com/zhangwx21/Object_Reconstruction/blob/master/structure_updated.png)



#### 数据集

ShapeNet rendered images http://cvgl.stanford.edu/data2/ShapeNetRendering.tgz

ShapeNet voxelized models http://cvgl.stanford.edu/data2/ShapeNetVox32.tg

#### 运行环境

python 3.5

tensorflow 1.13.0

numpy 1.13.3

scipy 0.19.0

matplotlib

skimage

PyMCubes

#### 训练模型

The code runs on HUAWEI's Modelarts. The main program is ./Code/main_AttSets.py


#### 其他说明

1.  Unzip the data set to ./Data/ShapeNetRendering and ./Data/ShapeNetVox32

2.  The trained model has been saved in ./Model/train_mod. You can use tensorboard to view the training process: ./Model/test_sum and ./Model/train_sum.
obs link:
obs://obsdeeplearning/3DOR/MA-Code-09-03/output/V0006/train_mod/
obs://obsdeeplearning/3DOR/MA-Code-09-03/output/V0006/test_sum/
obs://obsdeeplearning/3DOR/MA-Code-09-03/output/V0006/train_sum/

3.  The training log on Modelarts has been saved in the log folder.

4.  Due to the limitation of running time, the program only trained on the airplane data set(02691156).