Restricted Research and Evaluation License

Scope

This license applies to the model repository files, including model weights, metadata, calibration configuration, preprocessing configuration, model card text, and example inference materials.

Permitted Use

You may use the model and repository files only for:

private research;
educational review;
academic project evaluation;
non-commercial testing of the documented inference workflow.

Restrictions

You may not, without separate written permission from the copyright holders:

use the model or weights for commercial purposes;
redistribute, mirror, sublicense, sell, rent, or host the model weights;
use the model in production systems or public APIs;
use the model for legal, insurance, law-enforcement, safety-critical, or high-stakes decisions;
remove or obscure this license notice;
claim that the model is open source or approved for unrestricted reuse.

Training Data Notice

The model was trained on a multi-source vehicle image dataset assembled from Kaggle and Hugging Face datasets. Several upstream sources are marked as other or unknown license, and several image collections were assembled from web search engines or online communities. This license does not grant any rights to the original images beyond what their upstream owners and source platforms allow.

Before any broader public, commercial, or redistribution use, independently verify all upstream dataset licenses, source terms, attribution requirements, privacy obligations, and takedown requirements.

Model Architecture and Dependencies

The model uses a timm EfficientNetV2-S architecture with ImageNet-pretrained initialization and PyTorch-based inference. Third-party software remains governed by its own licenses.

No Warranty

The model and repository files are provided as-is, without warranties or guarantees of accuracy, safety, fitness for a particular purpose, non-infringement, or regulatory compliance.

Takedown and Contact

If you believe this repository contains content that violates a license, copyright, privacy right, or platform term, contact the maintainers so the material can be reviewed or removed.