W A L D O30

Introduction

WALDO (Whereabouts Ascertainment for Low-lying Detectable Objects) is an AI detection model utilizing a YOLO-v8 backbone and a synthetic data pipeline. It can detect various objects in overhead imagery, ranging from low-altitude to satellite images. WALDO supports multiple civilian applications such as disaster recovery, wildlife monitoring, and infrastructure oversight.

Architecture

WALDO is based on the YOLO-v8 architecture and is trained on custom datasets comprising synthetic and semi-synthetic data. The model outputs several object classes, including LightVehicle, Person, Building, UPole, Boat, Bike, Container, Truck, Gastank, Digger, Solarpanels, and Bus. It is designed for civilian use and does not include military applications.

Training

The WALDO model is trained using a unique dataset of synthetic and augmented data. Although the dataset is not publicly available, the model weights are open for deployment and further customization. Users can fine-tune the model on their data, setup sliding-window inference, or quantize the models for optimized performance on edge devices.

Guide: Running Locally

1. Clone the Repository: Access the code repository to get the boilerplate code necessary to run the models.
2. Set Up Environment: Install dependencies and ensure you have the Supervision annotation library from Roboflow.
3. Run the Model: Use the boilerplate code to execute the models and visualize detections.
4. Hardware Suggestions: For optimal performance, consider using cloud-based GPUs, such as those from AWS, Google Cloud, or Azure.

License

The WALDO model is released under the MIT License. This permits free use, modification, and distribution of the software, provided that the original copyright notice and permission are included in all copies. The software is provided "as is," with no warranties of any kind.