| | --- |
| | language: en |
| | license: agpl-3.0 |
| | library_name: ultralytics |
| | tags: |
| | - military |
| | - drone |
| | - object-detection |
| | - aerial-surveillance |
| | - defense |
| | metrics: |
| | - map |
| | - precision |
| | - recall |
| | model_name: VeritaMilitary |
| | base_model: |
| | - Ultralytics/YOLO26 |
| | base_model_relation: finetune |
| | --- |
| | |
| | # VeritaScan 🛡️ |
| | **VeritaScan** is a high-performance, lightweight object detection model specifically optimized for **Aerial Military Surveillance**. It is designed to be deployed on tactical drones with limited onboard computing power. |
| |
|
| | Developed by **M Mashhudur Rahim (XythicK)** at **Arkito Lab** (A Non-Profit Research Organization). |
| |
|
| | ## 🚀 Model Overview |
| | - **Model Size:** 5.4 MB |
| | - **Target Domain:** Aerial/Drone View |
| | - **Inference Speed:** ~2.7ms per image |
| |
|
| | ## 📊 Training Results |
| | The model was trained to achieve a solid balance between speed and accuracy for edge devices. |
| |
|
| | | Metric | Value | |
| | | :--- | :--- | |
| | | **mAP@.5-.95** | **88.7%** | |
| | | **Precision** | **96.2%** | |
| | | **Recall** | **98.6%** | |
| |
|
| | ### Performance Visualization |
| | The following graphs illustrate the training progress and model evaluation: |
| |
|
| | #### Training Curves |
| |  |
| |
|
| |
|
| | ## 🎯 Domain Specialization (Aerial vs Ground) |
| | **VeritaMilitary** is highly specialized for top-down perspectives. While it may show sensitivity to perspective shifts in ground-level imagery, it demonstrates exceptional precision in aerial views with confidence scores up to **90%**. |
| |
|
| | | Aerial Detection (Success) | | |
| | | :---: | :---: | |
| | |  |  | |
| | | **AirCraft Tracking** | | **Military Vehicle Tracking** | |
| |
|
| | ## 📦 Multi-Platform Deployment & Downloads |
| | VeritaMilitary supports a wide range of hardware. Click the buttons below to download the optimized formats: |
| |
|
| | | Framework | Target Hardware | Optimization | One-Click Download | |
| | | :--- | :--- | :--- | :--- | |
| | | <img src="https://files.svgcdn.io/simple-icons/onnx.png" width="40"> | **Universal** (PC, Cloud, Edge) | Standard FP32/FP16 | [**⬇️ Download ONNX**](https://huggingface.co/arkito/VeritaMilitary/resolve/main/veritascan.onnx) | |
| | | <img src="https://cdn.iconscout.com/icon/free/png-256/free-nvidia-logo-icon-svg-download-png-2945060.png" width="45"> | **NVIDIA Jetson / RTX** | CUDA Accelerated | [**⬇️ Download TensorRT (.engine)**](https://huggingface.co/arkito/VeritaMilitary/resolve/main/veritascan.engine) | |
| | | <img src="https://1000logos.net/wp-content/uploads/2021/05/Intel-logo.png" width="80"> | **Intel CPU / iGPU** | OpenVINO Runtime | [**⬇️ Download OpenVINO**](https://huggingface.co/arkito/VeritaMilitary/resolve/main/veritascan_openvino_model.zip) | |
| | | <img src="https://storage.googleapis.com/gweb-developer-goog-blog-assets/images_archive/original_images/image1_v7xhr8h.png" width="80"> | **Mobile / Android** | Mobile Quantized | [**⬇️ Download TFLite**](https://huggingface.co/arkito/VeritaMilitary/resolve/main/veritascan_float16.tflite) | |
| | | <img src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjuUv6hVLci1L1ZfvwIXDkZGetljXlYeUaVyDRyd5SxDf7HtGKrrpQY1Y4vrBWAK9lq9Ezmzw4Vk7v-dNeKyFCg6H_1Lqg43pGbNFHG5spM07j0ThWSPWGzkMdPg0P_KPsFO3-DHfKpR7I/s1600/tensorflowjs.png" width="80"> | **Web Browsers** | TensorFlow.js | [**⬇️ Download TF.js**](https://huggingface.co/arkito/VeritaMilitary/resolve/main/veritascan_web_model.zip) | |
| | | <img src="https://www.tensorflow.org/images/tf_logo_horizontal.png" width="80"> | **Legacy Systems** | Frozen Graph (.pb) | [**⬇️ Download TF-PB**](https://huggingface.co/arkito/VeritaMilitary/resolve/main/veritascan.pb) | |
| | | <img src="https://pytorch.org/assets/images/pytorch-logo.png" width="40"> | **C++ / Production** | TorchScript JIT | [**⬇️ Download TorchScript**](https://huggingface.co/arkito/VeritaMilitary/resolve/main/best.torchscript) | |
| | | <img src="https://cdn-icons-png.flaticon.com/512/0/747.png" width="35"> | **iOS / macOS** | Apple Neural Engine | [**⬇️ Download CoreML**](https://huggingface.co/arkito/VeritaMilitary/resolve/main/veritascan.mlpackage.zip) | |
| | | <img src="https://github.com/alibaba/MNN/raw/master/doc/banner.png" width="60"> | **Embedded Devices** | MNN Optimized | [**⬇️ Download MNN**](https://huggingface.co/arkito/VeritaMilitary/resolve/main/veritascan.mnn) | |
| | | <img src="https://raw.githubusercontent.com/Tencent/ncnn/master/images/256-ncnn.png" width="40"> | **Embedded Devices** | NCNN Optimized | [**⬇️ Download NCNN**](https://huggingface.co/arkito/VeritaMilitary/resolve/main/veritascan_ncnn_model.zip) | |
| | ## 🛠 Usage |
| | ```python |
| | from ultralytics import YOLO |
| | |
| | # Load VeritaMilitary |
| | model = YOLO('veritascan.pt') |
| | |
| | # Run Inference |
| | results = model.predict(source='drone_footage.mp4', imgsz=640, conf=0.25) |
| | results[0].show() |
| | ``` |
| | ### 🏢 About Arkito Lab |
| | Arkito Lab is a non-profit organization dedicated to open-source research and the development of AI solutions for humanitarian and defense technology. |
| |
|
| | Maintained by: M Mashhudur Rahim (XythicK) |
| | ### Cite This Project |
| | If you use this model in your research please cite |
| | ``` |
| | @software{XythicK_VeritaScan_2026, |
| | author = {M Mashhudur Rahim (XythicK)}, |
| | title = {VeritaScan: Lightweight Nano Object Detection for Aerial Military Surveillance}, |
| | year = {2026}, |
| | publisher = {Arkito Lab}, |
| | url = {https://huggingface.co/arkito/VeritaScan}, |
| | } |
| | ``` |
