AI Dog Breed Identifier
Upload a photo of a dog and get an instant AI-powered guess at its breed or breed mix. The photo is analyzed entirely in your browser using an on-device model — it's never uploaded to a server.
How to use it: upload a clear, well-lit photo with your dog's face and body visible, then click identify. A straight-on shot works better than an extreme angle or a photo where the dog is mostly obscured.
This is a fun estimate from a general-purpose image recognition model, not a DNA test — it recognizes around 120 common breeds and may not know rarer ones. Mixed-breed dogs often show several guesses with lower confidence, which usually reflects genuine visual similarity to more than one breed rather than an error.
How this works
This tool runs MobileNet, an open-source image classification model, directly in your browser using TensorFlow.js — the photo never leaves your device or touches a server. The model was trained on the ImageNet dataset, which includes roughly 120 common dog breeds, so recognition is strongest for well-established breeds and weaker for rare or very new ones. Want a full breakdown for a specific breed once you have a guess? Browse all breed guides.
How accurate is it?
Treat this as a fun, informed guess rather than a definitive answer — general-purpose image classifiers like MobileNet are strong at well-represented, visually distinct breeds and noticeably weaker at rare breeds, unusual coat colors, puppies (whose features haven't fully developed yet), and mixed-breed dogs that don't closely resemble any single trained category. A confident top match on a purebred-looking adult dog is usually reliable; a handful of similar-confidence guesses on an obviously mixed dog is the model honestly reflecting genuine visual ambiguity, not malfunctioning.
Frequently asked questions
Why does it show multiple breeds for one dog?
That's usually a sign your dog genuinely resembles more than one breed visually, or is a mix — the model returns its top predictions ranked by confidence, and for a mixed-breed dog it's common to see two or three plausible breeds with moderate, similar confidence rather than one clear high-confidence answer. That pattern is actually informative, not a failure of the tool.
Why doesn't it recognize my dog's breed?
The underlying model was trained on ImageNet, which includes roughly 120 common breeds — rarer, newer, or regionally-specific breeds outside that set simply aren't in its vocabulary, so it will guess the closest visual match among breeds it does know rather than say "unknown." If your dog is a less common breed, expect the guess to lean toward whichever recognized breed looks most similar.
Is my photo stored anywhere?
No — this runs entirely in your browser using TensorFlow.js, and the image never leaves your device or gets uploaded to a server. You could disconnect from the internet after the model finishes loading and it would still work, which is a good way to confirm nothing is being sent anywhere.
How is this different from a DNA test?
Completely different methods with very different reliability. This tool compares visual features (coat pattern, ear shape, body proportions) against what it learned from photos — it can be fooled by a mixed-breed dog that happens to look like a specific purebred, or miss a breed influence that doesn't show up visually. A DNA test reads actual genetic markers and is far more reliable for determining true ancestry, especially in mixed-breed dogs; this tool is a fun starting guess, not a substitute.