/3D-Object-Detection

Indoor Semantic Segmentation

Primary LanguagePythonMIT LicenseMIT

3D Object Detection

This repository is to do Indoor Semantic Segmentation with SegNet.

Dependencies

Dataset

image

Follow the instruction to download 2D-3D-S dataset.

$ wget https://storage.googleapis.com/3dsemantics/noXYZ/area_1_no_xyz.tar
$ wget https://storage.googleapis.com/3dsemantics/noXYZ/area_2_no_xyz.tar
$ wget https://storage.googleapis.com/3dsemantics/noXYZ/area_3_no_xyz.tar
$ wget https://storage.googleapis.com/3dsemantics/noXYZ/area_4_no_xyz.tar
$ wget https://storage.googleapis.com/3dsemantics/noXYZ/area_5a_no_xyz.tar
$ wget https://storage.googleapis.com/3dsemantics/noXYZ/area_5b_no_xyz.tar
$ wget https://storage.googleapis.com/3dsemantics/noXYZ/area_6_no_xyz.tar

Architecture

image

ImageNet Pretrained Models

Download VGG16 into models folder.

Usage

Data Pre-processing

Extract training images:

$ python pre-process.py

Train

$ python train.py

If you want to visualize during training, run in your terminal:

$ tensorboard --logdir path_to_current_dir/logs

Demo

$ python demo.py
Input GT Output
image image image
image image image
image image image
image image image
image image image
image image image
image image image
image image image
image image image
image image image

image