An UI tool designed for manually cropping two rectangular areas of an image in any customized size.
imglabeltool
├── dataset
│ └── xray
│ ├── 150057.jpg
| ├── 160022.jpg
| ├── 160565.jpg
| ├── 170116.jpg
| ├── 180006.jpg
| ├── annotation.csv <-- Most follow this format as an input of label.py
| └── coordinate.json <-- It's an output of label.py and the input of cut.py
├── cropped_L <-- Generate left desired region after running cut.py
│ ├──150057_L.jpg
│ :
│ └──
├── cropped_R <-- Generate right desired region after running cut.py
│ ├──150057_R.jpg
│ :
│ └──
├── cropped_L_flip <-- Generate mirrored left desired region after running cut.py
│ ├──150057_L_flip.jpg
│ :
│ └──
├── cropped_R_flip <-- Generate mirrored right desired region after running cut.py
│ ├──150057_R_flip.jpg
│ :
│ └──
├── cropped_cust_L <-- Generate left desired region in customed size after running cut.py -s xxx
│ ├──150057_L.jpg
│ :
│ └──
├── cropped_cust_R <-- Generate right desired region in customed size after running cut.py -s xxx
│ ├──150057_R.jpg
│ :
│ └──
├── cropped_cust_L_flip <-- Generate mirrored left desired region in customed size after running cut.py -s xxx
│ ├──150057_L_flip.jpg
│ :
│ └──
├── cropped_cust_R_flip <-- Generate mirrored right desired region in customed size after running cut.py -s xxx
│ ├──150057_R_flip.jpg
│ :
│ └──
├── cut.py
├── label.py
└── README.md
label.py -- Manually crop an image in a customized size. The coordinate, width and height are recorded in a json file.
cut.py -- Read the abovementioned json to genereate an image dataset.
-
Prepare the dataset and a csv. The csv should include all image ids.
-
Manually cropping.
$ cd imglabeltool
$ python label.py
- Generate cropped images in the size of 500x500.
$ python cut.py
Or if you want to generate a dataset in other size, e.g., 224x224 :
$ python cut.py -s 224