Utilization of this algorithm has proven to drastically reduce time, energy and cost required for image likeness comparison and recognition for still images and video frames. It can be used for:
• Whole or partial images
• Single or multiple objects
• Video content
• EXIF and MetaData
• Foreground, midground, background, or overlay
• AR / XR / VR content (including Lightstream)
• Spherical, volumetric, 360º and 3D (with built-in dewarping for wall, panoramic, ½ sphere, and full sphere)
• Low resolution to extremely high resolution (like logos, words, etc.)
• Common and professional / scientific image formats (HDRI)
The DejaVuAI® algorithm is the underlying engine for our Windows & Linux GUI application, Photon. This image recognition tool is able to identify likeness and look past images that have been altered by:
• Recoloring (including images taken with different sensors, like IR, thermal and scientific scanners, like terahertz)
• Arbitrary rotation
• Blurring
• Cropping (including irregular shapes)
• Partial Obscuring
• Scaling
• Noise (natural or artificial, including adversarial noise)
• Compression artifacts
• Skewing / shearing
• Flipping / mirroring
Search results currently provide:
• Visual comparison with automatic alignment and partial transparency
• Image registration (rotation angle, translation, shift, scale, mirror)
• Similarity Groups (shows how groups of images are related to each other, with multiple levels of association.)
• Confidence score (%)
• Multiple matches
