Dog Detective Bot

Inspiration

I wanted to experiment with computer vision and also telegram bots which prompted me to look into ways that I could do both

What it does

It is a Telegram bot that can predict the breeds of dogs in the images that you show it.

How I built it

I trained the model using Tensorflow Keras, along with VGG16, and was trained using the Stanford Dogs Dataset on Kaggle. The model was pre-trained before the hackathon.

I then used the model in the creation of the Telegram bot so that it can receive images as input and also output predictions, along with other simple Telegram bot functions.

Challenges we ran into

Sorting tens of thousands of images of dogs for model training was a nightmare, and coupled with my inexperience in computer vision model architecture made training the model a painful but fruitful experience.

I have also never coded a Telegram Bot before the hackathon started so learning and implementing it within a short span of time was extremely challenging.

Accomplishments that I'm proud of

I am proud of myself for completing this project as I did not expect myself to even come close to finishing and was just doing this for the learning experience. A great learning experience it was.