https://github.com/tanyachutani/unet-keras

Data-science-bowl-2018-Unet
https://github.com/tanyachutani/unet-keras

data-science-bowl-2018 image-segmentation keras segmentation tensorflow unet

Last synced: 21 days ago
JSON representation

Data-science-bowl-2018-Unet

• Host: GitHub
• URL: https://github.com/tanyachutani/unet-keras
• Owner: TanyaChutani
• Created: 2020-03-28T20:59:43.000Z (almost 5 years ago)
• Default Branch: master
• Last Pushed: 2020-04-24T09:07:42.000Z (almost 5 years ago)
• Last Synced: 2024-11-11T21:27:00.145Z (3 months ago)
• Topics: data-science-bowl-2018, image-segmentation, keras, segmentation, tensorflow, unet
• Language: Jupyter Notebook
• Size: 348 KB
• Stars: 1
• Watchers: 2
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Unet

Data-science-bowl-2018-Unet

## Data

The original dataset is data science bowl 2k18 - dataset (automating nucleus detection)  


and I've downloaded it, done the pre-processing and resizing all of the images to 128,128. 


Dataset link - [Dataset](https://www.kaggle.com/c/data-science-bowl-2018/data)

## Model

![](https://raw.githubusercontent.com/TanyaChutani/Unet/master/Architecture/u-net-architecture.png)


- Unet, deep neural network is implemented with Keras functional API.

- Output from the network is a 128*128 which represents mask that should be learned.

- The model was trained for 20 iterations with lr=0.001.

- Metric for the training is basically just a mean IOU.

## Results

You can download the pretrained weights from here  


[Weights](https://drive.google.com/open?id=10-A8XTgkr_xlDhsS1iclGCCjy_5Xs6xg)

Input Image. 


![](https://raw.githubusercontent.com/TanyaChutani/Unet/master/Output/unet1.png)


Mask. 


![](https://raw.githubusercontent.com/TanyaChutani/Unet/master/Output/unet2.png)