LKCell / docs /readmes /monuseg.md
qingke1's picture
initial commit
aea73e2
|
raw
history blame
1.12 kB
## MoNuSeg Preparation
The original PanNuke dataset has the following style using .xml annotations and .tiff files with a size of $1000 \times 1000$ pixels:
```bash
β”œβ”€β”€ testing
β”‚ β”œβ”€β”€ images
β”‚ β”‚ β”œβ”€β”€ TCGA-2Z-A9J9-01A-01-TS1.tif
β”‚ β”‚ β”œβ”€β”€ TCGA-44-2665-01B-06-BS6.tif
...
β”‚ └── labels
β”‚ β”œβ”€β”€ TCGA-2Z-A9J9-01A-01-TS1.xml
β”‚ β”œβ”€β”€ TCGA-44-2665-01B-06-BS6.xml
...
└── training
β”œβ”€β”€ images
└── labels
```
For our experiments, we resized the dataset images to $1024 \times 1024$ pixels and convert the .xml annotations to binary masks:
```bash
β”œβ”€β”€ testing
β”‚ β”œβ”€β”€ images
β”‚ β”‚ β”œβ”€β”€ TCGA-2Z-A9J9-01A-01-TS1.png
β”‚ β”‚ β”œβ”€β”€ TCGA-44-2665-01B-06-BS6.png
...
β”‚ └── labels
β”‚ β”‚ β”œβ”€β”€ TCGA-2Z-A9J9-01A-01-TS1.npy
β”‚ β”‚ β”œβ”€β”€ TCGA-44-2665-01B-06-BS6.npy
...
└── training
β”œβ”€β”€ images
└── labels
```
Everythin can be extracted using the [`cell_segmentation/datasets/prepare_monuseg.py`](cell_segmentation/datasets/prepare_monuseg.py) script.