Welcome to the repository for KINEVA, a selection of pre-trained computer vision models designed for a wide range of object detection tasks. This model is ideal for various applications, including complex scenes with challenging lighting and noise conditions, such as low-light security footage.
In this repository, we've compared KINEVA Model 1 with other well-known pre-trained models like YOLO 8 and YOLO 11. We want to demonstrate what you can do with different types of pre-trained models.
Below is an overview of the performance across various test images:
| Model | PARAMS | MAP | STATE | CATEGORIES | TYPE | VERSION |
|---|---|---|---|---|---|---|
| KINEVA Model 1 | 40M | 75.3 | RELEASED | HEAD PERSON NEGATIVES | VISION | 0.2B |
| KINEVA Model 2 | >80M | - | IN TRAINING | HEAD PERSON NEGATIVES BG | VISION | - |
This is our first model, use it at your own risk.
TBA (To Be Announced)
- Security
- Smart City
- Robotics
- Research
The model 1 is released in the models folder.
KINEVA Non-Commercial License (KNCL)
Version 1.0, October 2024
(See license.)
The public KINEVA Model 1 is trained on synthetic, public, and custom datasets. It contains the categories Person, Head, and Negatives. The additional background classes with a higher optimized model are not released yet.
We plan to release more open-source models, including additional versions of KINEVA. We are also working on integrating KINEVA with Ultralytics and training it on synthetic datasets to further improve its accuracy in different scenarios.
Feel free to explore and test KINEVA Model 1 in your own projects. You can also compare it with YOLO 8 and YOLO 11 for a better understanding of its capabilities.
Thank you for trying out KINEVA Model 1! We welcome any feedback or contributions. 😊
- Python 3.9 or higher
- PyTorch 1.10 or higher
- Other dependencies (see
requirements.txt)
-
Clone the repository:
git clone https://github.com/your-repo/kineva.git
-
Install the necessary dependencies:
pip install -r requirements.txt
-
Download the KINEVA Model 1 from the models folder.
-
You're ready to go! You can start using the model in your projects.
To run inference with KINEVA Model 1, you can use the following example:
import torch
from kineva import KinevaModel
# Load the model
model = KinevaModel('path_to_kineva_model')
# Load an image
image = 'path_to_image.jpg'
# Run inference
results = model.predict(image)
# Display the results
print(results)For further examples and detailed explanations, refer to the examples folder.
We welcome contributions to the KINEVA project! Please follow these steps:
- Fork the repository.
- Create a new branch for your feature or bug fix.
- Submit a pull request with a clear description of the changes.
- Ensure all tests pass before submitting your PR.
For major changes, please open an issue first to discuss what you would like to change.
If you have any questions, suggestions, or issues, feel free to open an issue on GitHub or reach out to us at:
- Page: @kineva_ai
- LinkedIn: @kineva_ai