🐶 About DogLens
AI-powered dog breed identification built to help dog owners, prospective adopters, and dog lovers make informed, responsible decisions
🎯 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.
AI Breed Identification
Upload a photo and our machine learning model analyses it to identify the breed, providing confidence scores and detailed breed composition breakdowns.
Breed Characteristics
Detailed profiles covering size, energy levels, grooming needs, trainability, health predispositions, and regional cost estimates across the UK.
Training & Behaviour
Guides on training fundamentals, breaking bad habits, understanding canine communication, managing anxiety, and building a strong bond with your dog.
Health & Wellness
UK-specific veterinary information including vaccination schedules, parasite prevention, dental care, common illnesses, and when to see a vet.
Shelter & Breeder Finder
Tools to help you find reputable rescue centres and responsible breeders near you, because where you get your dog matters as much as which breed you choose.
Lifestyle Compatibility
An interactive quiz that scores how well a breed matches your living situation, activity level, experience, household, and available time.
🧠 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.
| Component | Technology | Purpose |
|---|---|---|
| ML Framework | Microsoft ML.NET | Manages the training pipeline, model loading, and predictions |
| Neural Network | ResNetV2-101 | 101-layer deep residual network for image feature extraction |
| Training Technique | Transfer Learning | Adapts pre-trained ImageNet knowledge to dog breed classification |
| Pre-Training Data | ImageNet (1.2M images) | Provides foundational visual understanding of shapes, textures, patterns |
| API Server | ASP.NET Core | Hosts the model and serves predictions via REST API |
| Website | HTML, CSS, JavaScript | User interface for uploads, results display, and educational content |
⚖️ 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.
⚠️ Important: DogLens provides its best estimate based on visual analysis. It is not a DNA test and should not be treated as a definitive breed determination. For official breed identification, a canine DNA test from a provider such as Wisdom Panel or Embark is recommended.
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 is continuously expanded, with pages on health, behaviour, travel, law, and more being researched and written from authoritative UK sources including the PDSA, Blue Cross, Kennel Club, Dogs Trust, and GOV.UK.
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. This is a genuine labour of love for dogs and technology.
Future plans include expanding the breed database, adding more educational content, building community features, 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.