We came across the idea of identifying dog breeds when we realize that many of our friends were unable to find the breed of a dog whenever they saw one on the streets. We came to the realization that the use of machine learning and neural networks can be a viable solution to this issue.
What it does
DoggoBot is a Telebot incorporated with a TL-CNN model that identifies a dog's breed based on a photo of it. By submitting a photo of a dog, DoggoBot will process the photo and classify the dog's breed by running through the saved model and predicting its classification.
How we built it
We made use of mainly Tensorflow for the neural network building and TelegramBot for the TeleBot.
Challenges we ran into
Reading docs, debugging, using unfamiliar libraries and formats
Accomplishments that we're proud of
That we managed to make a product that actually runs and a model that has a fairly high accuracy and (and potential for higher given more time)
What we learned
TensorFlow, transfer learning, CNN, telegram bots, webhooks, heroku, git
What's next for DoggoBot
More accurate machine predictions!