jaimead7-vimi 0.0.2

Interface for AI vision models.

VIMI - Vision Models Interface

---

Interface for AI vision models.

Authors

Jaime Alvarez Diaz

Installation

Install as a package from source files

git clone https://github.com/Jaimead7/ViMI.git
cd ViMI
py -m pip install .

Install as a package from pypi

py -m pip install jaimead7-vimi

Usage

Create a model folder with a valid structure. You can find examples in ./test/models/.

from pathlib import Path
from typing import Optional

import cv2
import numpy as np

from vimi.engines import EnginesReg, ModelEngine
from vimi.results import Results


MODEL_PATH: Path = Path('./test/models/DetectModelNCNN')
IMGS_PATHS: list[Path] = [
    Path(Path('./test/imgs/coco/bike.jpg')),
    Path(Path('./test/imgs/coco/planes.jpg'))
]

model: Optional[ModelEngine] = EnginesReg.get_model(MODEL_PATH)
if model is None:
    raise RuntimeError(f'Cant\'t open the model "{MODEL_PATH}".')

results: list[Results] = model(IMGS_PATHS)

if len(results) != len(IMGS_PATHS):
    raise RuntimeError(f'Error processing images: Expected {len(IMGS_PATHS)} Results, got {len(results)}.')
for result, img in zip(results, IMGS_PATHS):
    cv2.imshow(f'{img.name}', result.plot())
_: int = cv2.waitKey(0)
cv2.destroyAllWindows()

Available model interfaces

YOLO11 models:

type	ncnn	pyTorch
classify	✅	❌
detect	✅	❌
obb	✅	❌
pose	❌	❌
segmentation	❌	❌

License

This project is licensed under the MIT license.

Third-party models:
The ./test/ directory contains model files that are subject to their own respective licenses. Each model includes a metadata.yaml file specifying its applicable license terms. Please refer to these metadata files for details.

Contributing

Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.

Please make sure to update tests as appropriate.