teeth_segmentation

teeth segmentation using UNet and customize attention module

https://github.com/saeedahmadicp/teeth_segmentation

Science Score: 44.0%

This score indicates how likely this project is to be science-related based on various indicators:

  • CITATION.cff file
    Found CITATION.cff file
  • codemeta.json file
    Found codemeta.json file
  • .zenodo.json file
    Found .zenodo.json file
  • DOI references
  • Academic publication links
  • Academic email domains
  • Institutional organization owner
  • JOSS paper metadata
  • Scientific vocabulary similarity
    Low similarity (9.4%) to scientific vocabulary

Keywords

communityexchange dentistry educative learn medical-imaging pytorch unet-image-segmentation
Last synced: 4 months ago · JSON representation ·

Repository

teeth segmentation using UNet and customize attention module

Basic Info
  • Host: GitHub
  • Owner: saeedahmadicp
  • License: mit
  • Language: Python
  • Default Branch: master
  • Homepage:
  • Size: 254 MB
Statistics
  • Stars: 26
  • Watchers: 3
  • Forks: 3
  • Open Issues: 0
  • Releases: 0
Topics
communityexchange dentistry educative learn medical-imaging pytorch unet-image-segmentation
Created about 3 years ago · Last pushed almost 2 years ago
Metadata Files
Readme Contributing License Citation

README.md

Teeth Segmentation Using UNet and Its Variants

This repository contains the code for training and evaluating various UNet models for teeth segmentation. The models implemented include the original UNet, as well as some of its variants such as UNet++, ResUNet, and Attention UNet.

Data

The data used for this project is not publicly available, but you can request it by contacting me through the email address provided on the profile page. Once you have the data, make sure to update the paths accordingly.

Usage

Before running the code, make sure to modify the 'train.py' file and the other relevent files so that you can get the relevent results.

To train the model, simply run the train.py script:

bash python train.py

Results for the test data

| UNet Variants | Test Accurary | Test Dice Score | |----------|----------|----------| | Base UNet | 96.10 | 90.47 | | UNet with GN | 96.71 | 91.53 | | Attention UNet | 96.40 | 91.01 | | Spatial Attention UNet | 96.45 | 91.09 | | Inception UNet | 96.29 | 90.69 | | Residual UNet | 96.16 | 90.06 | | UNet++ | 96.11 | 90.33 | | Dense UNet with GN | 96.77 | 91.88 | | Spatial Attention UNet2 ${\color{red}^*}$ | 97.32 | 93.12 |

${\color{red}*}$ increase the resolution from 256*256 to 768*512, reduce the batch size from 16 to 2, used Group Normalization and Custom spatial attention module with base UNet

${\color{red}Note}$ This project is solely for learning purposes; no standard practices are applied. Therefore, I am not claiming any state-of-the-art results.

Owner

  • Name: Saeed Ahmad
  • Login: saeedahmadicp
  • Kind: user
  • Location: Geumcheon District, Seoul, South Korea
  • Company: IKLab Inc.

AI Research Engineer @ IKLab Inc. | Machine Learning Engineer

Citation (CITATION.cff)

authors:
  - family-names: Ahmad
    given-names: Saeed
message: "If you use this software, please cite it using these metadata."
title: "Teeth Segmentation using PyTorch"

GitHub Events

Total
  • Watch event: 5
  • Fork event: 1
Last Year
  • Watch event: 5
  • Fork event: 1