# CodeReview

**Repository Path**: ysl9707/code-review

## Basic Information

- **Project Name**: CodeReview
- **Description**: EI-UNet和改进型U-Net在ISIC2018数据集上实验
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2024-08-15
- **Last Updated**: 2024-08-20

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

## 比较EI-UNet与一种改进型U-Net

### Data preparation

We cropped the ISIC 2018 dataset to 224x320 and 256x256,both saved it in npy format,  which can be downloaded from Baidu web disk. 

```
link: https://pan.baidu.com/s/1bIVUdzYG_7tuwalbI4Y8Ww

password: c36c

link: https://gitee.com/ysl9707/code-review
```

Place the downloaded npy files in the "data" directory and unzip them. The decompression format is as follows:

```
/data/ISIC2018_npy_all_224_320/image/

​		ISIC_0000000.npy

​		ISIC_0000001.npy

​		...

​		ISIC_0016072.npy

/data/ISIC2018_npy_all_224_320/label/

​		ISIC_0000000_segmentation.npy

​		ISIC_0000001_segmentation.npy

​		......

​		ISIC_0016072_segmentation.npy
```

### Train and Test

only run main_train.py, you can train four Models(Ms_red、EIU_Net、UNext、UNet(improved)) 


```
python main_train.py
```

### Performance on ISIC 2018
on 224_320

| Networks  |     JI     |    Dice    |  ACC   |   SE   |   SP   |
|-----------|:----------:|:----------:|:------:|:------:|:------:|
| UNeXt     |   0.8170   |   0.8865   | 0.9706 | 0.8923 | 0.9849 |
| Ms RED    |   0.8291   |   0.8979   | 0.9608 | 0.9044 | 0.9679 |
| EI-UNet   |   0.8316   |   0.8968   | 0.9637 | 0.9069 | 0.9769 |
| UNet(our） | **0.8363** | **0.8997** | 0.9631 | 0.9032 | 0.9704 |

on 256_256(opt=Adam lr=1e-3 no scheduler)

| Networks  |     JI     |    Dice    |    ACC     |   SE   |   SP   |
|-----------|:----------:|:----------:| :--------: |:------:|:------:|
| UNeXt     |   0.7676   |   0.8524   |   0.9373   | 0.9080 | 0.9604 |
| Ms RED    |   0.7837   |   0.8673   |   0.9228   | 0.9341 | 0.9273 |
| EI-UNet   |   0.7512   |   0.8265   |   0.9273   | 0.8106 | 0.9867 |
| UNet(our） | **0.7984** | **0.8748** | **0.9439** | 0.9250 | 0.9581 |

### Note
folder1——Ms RED source code dataset split  
folder10——EI-UNet source code dataset split  
folder11——ISIC2018 raw dataset split