Blackshark ORCA™ HUNTR No-Code Object Detection

ORCA HUNTR - OBJECT DETECTION AND CLASSIFICATION IN MINUTES ORCA HUNTR is a revolutionary no-code environment that enables anyone to train, test and exploit AI/ML-powered object detection and classification, which can be applied to a variety of scenarios.

INTUITIVE OBJECT DETECTION AND CLASSIFICATION: A simple user interface that requires no knowledge of AI or coding, enables an analyst to quickly train the AI model to detect and classify objects in up to 10 different categories from any RGB image, whether it was acquired by satellite, aerial or other photographic methods. Using it is as simple as coloring objects with a crayon or a paint brush.

RAPID DEVELOPMENT, IMMEDIATE RESULTS: Real-time feedback from the AI model enables the analyst to quickly visualize results and where additional labeling input is needed. The model can then be used to detect results in larger areas with a click of a button. Compared to conventional AI development process, ORCA HUNTR can build a fully functional detection model at very low cost, in minutes or hours as opposed to weeks and months. Detected results can be exported from the platform with georeferenced vectors easily integrated into downstream GIS systems or applications. Fully trained models can be packaged to be deployed in any execution environment: cloud, edge or fully embedded.

MULTI-SENSOR, MULTI-MODE: ORCA HUNTR can process any RGB image regardless of resolution and acquisition method and up to 15 channels. From satellite to aerial or drone-captured imagery to JPEG files captured by a digital camera or still frames from a video camera, ORCA HUNTR can deliver insights about whether objects are present, or segment them into categories and provide centroids and bounding boxes of objects in a scene.

SAFE & EXPLAINABLE AI: The training process always involves a human-in-the-loop approach and at every step, for every label applied, the human receives immediate real-time feedback from the model as it is being trained. The operator has access to detailed information about the model predictions and also its uncertainty (model entropy) via simple visualizations, enabling the operator to quickly address areas or objects of high uncertainty in the training process.