Advanced Image Matching
DejaVuAI is our index-driven image search technology – an unsupervised algorithmic image search technology that can be utilized locally or in a distributed environment. It is 100% image-based searching – no keywords or metadata required. With as little as .5% of a parent image, users can rapidly identify images and videos in a sea of noise.
DejaVuAI is immediately acquirable via the CDAO Tradewinds Solutions Marketplace.
See Once, Remember Forever™
Identify humans, uniforms,
equipment, and vehicles fast
With only part of an image, DejaVuAI can identify people, uniforms, hardware, and physical locations. Find other images a person is included in, despite being in a crowd of people or wearing something that is blocking part of their face, like a mask or a hat. It can be skewed, flipped, cropped, even with color distortions, adversarial noise, and partial obstructions, yet we can still find matches. With just a basic laptop, DejaVu can identify people and things in seconds.
See through the noise
DejaVuAI allows image-based searching regardless of:
-
Cropping (including irregular shapes)
-
Recoloring (including images taken with different sensors, like IR, thermal, and scientific scanners)
-
Blurring or noise (optical or filter applied)
-
Partial obscuring
-
Skewing / shearing / scaling
-
Arbitrary rotation
-
Flipping / mirroring
-
Natural or applied compression artifacts
-
Partially transparent overlay
AI Change Detection
Find out how many images of a specific vehicle can be found, and see where it's been. This can be used for cars, trucks, motorcycles, planes, trains, tanks, ships, and any other type of vehicle you can think of. Even cloud formations can be matched. Use our similarity groups to see what images have the closest resemblance to the original. DejaVuAI can find the important identifying information you're looking for in seconds.
Unlike a human, DejaVu AI models remember the precise configuration of military vehicles forever, and can instantly recall past configurations and compare them to current loadouts.
Explainable AI Models
Understand how your results match with our similarity groups, confidence percentages, and a slider which will visually show you how two images match, giving you an extra level of control and certainty in our system.
• 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)