SwiftLoader

SwiftLoader is a set of dataloaders opotimized for pytorch.

Features

• Classification: The classification.py module provides a dataloader for classification tasks.
• Object Detection: The object_detection.py module provides a dataloader for object detection tasks.
• Template Based Detection: The template_based_detection.py module provides a dataloader for template based detection tasks.

Installation

To install SwiftLoader, run the following command inside the root directory of the repository:

pip install -e .

Usage

To use the provided dataloaders the dataset should have the following structure:

root
├── dataset_1
|   ├── categories.json
│   ├── scene_1
│   │   ├── images
│   │   │   ├── 000000.jpg
│   │   │   ├── 000001.jpg
│   │   │   ├── ...
│   │   ├── annotations
│   │   │   ├── 000000.json
│   │   │   ├── 000001.json
│   │   │   ├── ...
│   │   ├── image_annotations
│   │   │   ├── 000000.jpg
│   │   │   ├── 000001.jpg
│   │   │   ├── ...
...
├── objects

The categories.json file should contain a list of categories in the dataset. For example:

[
    {"id": 0, "name": "category_1", "supercategory": "category_1"},
    {"id": 1, "name": "category_2", "supercategory": "category_2"},
    ...
]

Image annotation files should contain the following fields: id, width, height, file_name. Any additinal annotations should be put in the "attributes" field. For example:

{
    "id": 0,
    "width": 1000,
    "height": 1000,
    "file_name": "000000.jpg",
    "attributes": {
        "attribute_1": "value_1",
        "attribute_2": "value_2",
        ...
    }
}

Object detection annotation files should contain the following fields: image_id, category_id, bbox. Any additinal annotations should be put in the "attributes" field. For example:

[
    {"image_id": 0, "category_id": 0, "bbox": [0, 0, 100, 100], "attributes": {"attribute_1": "value_1", "attribute_2": "value_2", ...}},
    {"image_id": 0, "category_id": 1, "bbox": [100, 100, 200, 200], "attributes": {"attribute_1": "value_1", "attribute_2": "value_2", ...}},
    ...
]

To use SwiftLoader, import the necessary modules from the swiftloader package. For example:

```python
from swiftloader import SwiftClassification
from swiftloader import SwiftObjectDetection
from swiftloader import SwiftTemplateObjectDetection

## License
This project is licensed under the MIT License - see the (LICENSE) file for details.