🐶 About DogLens

🎯 Our Mission

DogLens was created with a simple but important goal: to help people understand dogs better. Whether you're considering getting your first dog, trying to identify a breed you've spotted on a walk, or wanting to learn more about your own dog's needs, DogLens is here to help.

Too often, dogs are chosen based on appearance alone. A cute puppy in a window or an adorable photo online doesn't tell you that breed needs two hours of exercise a day, or that it's prone to separation anxiety, or that its grooming costs hundreds of pounds a year. DogLens puts all of that information at your fingertips, alongside AI-powered breed identification so you know exactly what you're looking at.

🛠️ What DogLens Does

DogLens is more than a breed identifier. It's a comprehensive resource designed to support every stage of dog ownership, from the initial spark of interest through to caring for a senior dog in their twilight years.

❓ How Ask a Question Works

The Ask a Question page lets you type any question about dogs in your own words, such as “How dangerous is a tooth abscess?”, “Can grapes make my dog ill?” or “Does a joint supplement actually work?”. The aim is to give you a proper answer, in enough detail that you don't need to go anywhere else, while always showing you where it came from so you can check it.

Where the Answers Come From

The DogLens guides always come first. When they cover your question, the answer comes from them, and when the best answer is on another site, our own page on the subject is still offered so you can read more here. Beyond our guides, the page answers from facts learnt from trusted sources, such as Wikipedia, GOV.UK and published veterinary research. Each fact is only used after it has been checked against the page it came from, and each one is shown with a link to that page so you can read the original for yourself. Nothing is made up: if there is no reliable answer yet, the page says so honestly rather than guessing.

“Is It Proven?”

Many questions are really about whether something works: a supplement, a food, a remedy or a product. For these, the page shows what published studies in dogs actually found, and how they tested it. When there are no study findings to show, it tells you plainly, so a maker's claims are never presented as proof.

What It Will Never Do

Always Learning

Questions the page can't answer yet are noted (only the question itself is kept), so it knows what people want to know. In the background, DogLens then researches those subjects from trusted sources, checks the facts and adds them, so the page can answer more over time, including things not yet covered in our own guides. All of this runs on our own computer, using our own AI models. Facts already in use are rechecked against their sources about every six months, so they stay up to date.

🧠 How the AI Works

The breed identification at the heart of DogLens is powered by machine learning, a branch of artificial intelligence where a computer learns to recognise patterns by studying thousands of examples rather than following hand-written rules. Here's what happens when you upload a photo.

The Technology Behind It

DogLens uses Microsoft ML.NET, an open-source machine learning framework, combined with a deep neural network architecture called ResNetV2-101 (Residual Network, Version 2, 101 layers deep). This is the same family of technology used by researchers at Microsoft and in academic computer vision labs around the world.

ResNetV2-101 was originally trained on ImageNet, a massive dataset of over 1.2 million images across 1,000 categories. This pre-training gives the network a deep understanding of visual concepts like edges, shapes, textures, and patterns. DogLens then uses a technique called transfer learning to take that general visual knowledge and specialise it for dog breed recognition.

What Is Transfer Learning?

Training a neural network from scratch would require millions of dog images and weeks of computing time. Transfer learning is a much smarter approach. Think of it like this: rather than teaching someone to see from birth, you take a person who already understands shapes, colours, and textures, and simply teach them the specific differences between a Labrador's broad head and a Greyhound's narrow one.

In technical terms, we take the pre-trained ResNetV2-101 model (which already understands general visual features) and retrain only the final classification layers using our curated dataset of dog breed images. The earlier layers, which detect edges, textures, fur patterns, and shapes, remain frozen. The result is a model that achieves strong accuracy with far less training data than would otherwise be needed.

How the Model Learns Dog Breeds

The training process works in stages. First, hundreds of images are collected for each breed from carefully vetted sources, ensuring variety in angles, lighting, backgrounds, and individual dogs within each breed. These images are then fed through the network in batches. During each training cycle (called an epoch), the model makes predictions, compares them against the known correct answers, and adjusts its internal parameters to reduce errors. Over many epochs, the model gradually learns the subtle visual differences that distinguish one breed from another.

After training, the model is packaged into a single file that the DogLens API loads when the server starts. When you upload a photo, the image passes through all 101 layers of the network in a fraction of a second, and the model outputs a confidence score for every breed it knows. The highest-scoring breeds are returned as the prediction.

⚖️ Accuracy & Limitations

DogLens is an experimental tool and we believe in being upfront about what it can and can't do. The AI is continuously improving as more breeds and training images are added, but there are inherent limitations to any image-based identification system.

The model performs best with clear, well-lit photographs showing the full body or a clear head-on view of the dog. Factors that can affect accuracy include poor lighting, unusual angles, heavy grooming or styling that obscures natural features, very young puppies whose breed characteristics haven't fully developed, and mixed-breed dogs where features from multiple breeds are present.

Mixed-breed dogs are particularly challenging for any visual identification system. DogLens will show the breeds it detects with confidence percentages, which can give you an indication of the likely mix, but this is an educated estimate rather than a genetic analysis.

📊 DogLens at a Glance

The breed count grows as new breeds are trained and added. Check the main page for the current total.

🌱 Always Growing

DogLens is an active, evolving project. New breeds are regularly added to the AI model as training images are collected and validated. The educational content now spans over 60 pages covering health and veterinary care, training and behaviour, home and family life, travel and workplaces, dog science and cognition, world cultures, UK law, and much more — all researched and written from authoritative UK sources including the PDSA, Blue Cross, Kennel Club, Dogs Trust, RSPCA, GOV.UK, and peer-reviewed veterinary and animal behaviour journals.

Every aspect of DogLens is built and maintained by a single developer based in Milton Keynes. The AI model is trained locally, the API is self-hosted, and every information page is researched, written, and coded by hand. There are no auto-generated articles or content scraped from other sites. The Ask a Question page is the one place that also shows facts from other trusted websites, and every one of them is credited and linked to its source. This is a genuine labour of love for dogs and technology.

The Bark Board community message board is live — dog owners can share stories, tips, and thoughts with the DogLens community. Recent additions include pages on working dogs, dogs around the world, how breeds are created, and why some dogs become anti-social. Future plans include expanding the breed database, adding more educational content, and eventually deploying a mobile app so you can identify breeds on the go during walks and park visits.

👋 Who Made This

DogLens was created by Clive Homewood, a developer based in Milton Keynes, United Kingdom. What started as a personal project to explore machine learning quickly grew into a comprehensive resource for dog lovers. The combination of AI technology and genuine passion for helping dogs and their owners find the right match has shaped everything you see on this site.

If you have questions, suggestions, or would like to get in touch, please visit the Contact page.