Contact Us
Back to all case studies

Dog Face Recognition and Identification for a Mobile App

Dog face identification accuracy – 91.96%.

Dog Face Recognition
Key Details

Dog face identification accuracy – 91.96%.

  • Challenge
    Dog face identification and breed recognition
  • Solution
    A deep learning approach for dog face verification
  • Technologies and tools
    FasterRCNN, ArcFace, ResNet50


Pet ownership is on the rise, as well as the number of lost pets is. The American Humane Association states that 1 out of 3 pets become lost at some point of their lifetime. With the numbers rising up, there’s a need for dog face detection and identification using modern visual technology. Today, facial recognition is used for locating lost pets.

The client is a software development company. They were looking for a trusted face identification development company. They needed an ML algorithm for dog face recognition and breed classification for a mobile app.

The mobile app will be a pet social network to connect pets and their owners. Key features of the app:

  • create a pet profile
  • upload and store your pet’s photos
  • find other pet profiles near you
  • face recognition dog matching
  • face recognition search dog if lost

Challenge: dog face identificationand breed recognition

Dogs of the same breed look practically the same. Many dogs of one particular breed can be difficult to tell apart. In case there are several listings of Alaskan Malamute in the app, an algorithm struggles to recognize the faces of dogs without good facial recognition technology.

To solve this problem, our data scientists collected datasets of dog breed images. We used them for supervising the training of learning algorithms as well as testing them.

Solution: a deep learning approach for dog face verification

dog face recognition matches

We built a dog face recognition and breed identification module for the client’s pet app. The module included two models for pet facial recognition and breed classification. Deep learning-based models use images of a found pet and match them against a database of images from all pet profiles, to find a match instantly.

The lost pet’s owner doesn’t have to look through tons of listings, the system suggests matches based on image analysis. If there’s a match, the user can contact the person’s found their pet. This works both ways: the person who found a lost pet takes a photo and gets a match if someone is looking for this pet.

We used FasterRCNN for the face detection of dogs. ArcFace for face identification, and ResNet50 for breed classification. There were over 130 breeds analyzed.

Dog face identification accuracy – 91.96%

Breed classification accuracy – 83.29%

Our Approach:

  1. A user uploads a new picture of a pet in the application.
  2. Verification of whether a pet face is present on the image.
  3. Verification of the lighting, image quality, facial features verification
  4. NN using dogface shapes performs breed identification
  5. If no similar pet exists within the database, the system assumes it’s a new pet, otherwise, returns a match of possible candidates
  6. ID assignment to the pet, and their landmarks for further storage.

Our team has also developed an API connected to the neural network for smooth frontend and backend interaction.

Result: dog face recognition and breed identification

We’ve developed a neural network for face detection of dogs, recognition and verification. Our DL-based solution allows identifying a dog through one or a series of pictures given by a user, based on their facial features. This solution is the key to finding lost dogs more effectively. It helps connect lost pets to their owners and bring them home faster.

Let’s Develop a Pet Face Recognition App with InData Labs

  • Computer Vision

Contact InData Labs

Want to start getting value from your data? Fill the form. Click send. Let's talk.

    By clicking Send Message, you agree to our Terms of Use and Privacy Policy.