SnapSort can help you identify & sort images into different directories. It uses OpenAI's CLIP model locally. No internet connection required.
- Automated Image Classification: Leverages CLIP's deep learning model for precise image recognition.
- Custom Sorting Labels: Allows flexibility in sorting images into user-defined categories.
- Batch Processing: Efficiently handles large volumes of images.
- Dry Run Mode: Test sorting criteria without moving files to ensure precision.
- Simple CLI: Easy-to-use command-line interface for seamless operation.
SnapSort can be installed using poetry. To install, clone the repository and run the following commands:
poetry shell
poetry install
python snapsort.py --dir /path/to/your/images
The default labels are for a use-case that detects screenshots & receipts and sorts them into separate folders. Any other images are left in the original directory.
[
"a screenshot of a software interface or a screen capture from phone",
"a photo of an invoice or a receipt",
"a photo of a real-world scene, an object, a person, or any image not fitting the description of a screenshot, receipt, or invoice",
]The last label should be a catch-all label for images that do not match any of the other labels.
The threshold is the minimum confidence score required for an image to be sorted into a category. You need to experiment with this value to find the best threshold for your use case. Set it with --threshold flag. The default value is 0.5.
To dry run without moving files, add the --dry-run flag.
Set an appropriate batch size with the --batch-size flag. The default value is 4.
python snapsort.py demo/ -t 0.5 --dry-run
| file | class | probability | status | |
|---|---|---|---|---|
| 0 |  |
a screenshot of a software interface or a screen capture from phone | 0.98 | Moved (Dry Run) |
| 1 |  |
a photo of an invoice or a receipt | 0.97 | Moved (Dry Run) |
| 2 |  |
a photo of a real-world scene, an object, a person, or any image not fitting the description of a screenshot, receipt, or invoice | Skipping | |
| 3 |  |
a screenshot of a software interface or a screen capture from phone | 0.76 | Moved (Dry Run) |
| 4 |  |
a photo of an invoice or a receipt | 0.99 | Moved (Dry Run) |
| 5 |  |
a photo of a real-world scene, an object, a person, or any image not fitting the description of a screenshot, receipt, or invoice | Skipping | |
| 6 |  |
a screenshot of a software interface or a screen capture from phone | 0.63 | Moved (Dry Run) |
| 7 |  |
a photo of a real-world scene, an object, a person, or any image not fitting the description of a screenshot, receipt, or invoice | Skipping | |
| 8 |  |
a screenshot of a software interface or a screen capture from phone | 0.76 | Moved (Dry Run) |
| 9 |  |
a photo of a real-world scene, an object, a person, or any image not fitting the description of a screenshot, receipt, or invoice | Skipping | |
| 10 |  |
a photo of a real-world scene, an object, a person, or any image not fitting the description of a screenshot, receipt, or invoice | Skipping | |
| 11 |  |
a photo of a real-world scene, an object, a person, or any image not fitting the description of a screenshot, receipt, or invoice | Skipping |
- Python 3.6+
- PyTorch
- PIL (Python Imaging Library)
- transformers by HuggingFace
We welcome contributions to SnapSort. Feel free to open a pull request or issue.
SnapSort is made available under the MIT License.