PawSpective, an innovative browser experiment, uses AI and colour transformation to simulate how dogs see their environment, highlighting differences in visual perception and fostering responsible product design.
PawSpective is a browser-based experiment that starts from a simple but familiar problem: a toy that is easy for a person to spot may be much harder for a dog to notice. The project is built around the well-established fact that dogs do not see colour in the same way humans do. Veterinary and canine-focused sources such as PetMD and the American Kennel Club describe dog vision as dichromatic, with stronger sensitivity to blues and yellows and much weaker distinction between reds and greens.
Rather than presenting itself as a scientific model of canine sight, the application frames its output as an approximation. Its live “Dog Lens” applies a colour transformation in the browser, then lets the user compare a human view with a canine-inspired version. That approach fits the broader science on dog vision, which also points to a different visual balance from human sight and a greater dependence on motion and contrast cues, as Scientific American and PetMD have explained.
The more technical part of the system combines video upload, AI scene analysis and deterministic image processing. A short clip is sent to Google Gemini, which returns structured labels, bounding boxes, timestamps and confidence data. The user can then correct those detections before the app measures how strongly a selected object stands out from its background. That distinction matters: PawSpective treats the AI as an interpreter, not as the source of a visibility score.
The visibility and “curiosity” features are calculated with OpenCV, NumPy and colour-space conversions rather than model confidence alone. The app samples foreground and background pixels, applies the same canine-style transformation used in the preview, then compares contrast in CIE Lab terms. It also tests alternative toy colours against the same background, which is a direct nod to the practical advice often given by canine organisations: blue and yellow objects tend to be easier for dogs to distinguish than red or green ones.
PawSpective ends with a fictional story reel, produced from the reviewed scene data and presented explicitly as entertainment. The workflow uses Gemini for grounded scene interpretation, ElevenLabs for narration and FFmpeg for final composition. The maker’s larger point is clear: playful AI products can still be careful about boundaries, especially when they separate what is measured, what is inferred and what is invented. That makes the project more than a novelty; it is a small case study in responsible product design.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
