💡 Inspiration
I am sure everyone who has ever watched the TV series Silicon Valley remembers Jian Yang's HotDog/Not HotDog classifier. We thought it would be really cool if we tried to build a shark version of the same for the hackathon. It is modeled after the HotDog app and also has a login screen which involves a meme from the series.
🏡 What it does
When the phone camera is pointed at something while in the app, it classifies that as Shark/Not Shark (Hence the name!).
🔨 How we built it
There are two major components:
- The mobile app: For this we used Flutter so that we could build the app for both Android and iOS. To then classify images taken using the camera, we used the
tflitelibrary and used it to interface the model and the app. - Shark Classifier: We built an ML model using TensorFlow and Keras that classifies images and tells you whether an image is a shark or not. This is then converted to a TensorFlow Lite model to run on mobile.
🥲 Challenges we ran into
- Gathering and organizing the dataset for Sharks and Not Shark objects.
- Interfacing the Tensorflow model with the flutter app - omg was this difficult. We had so many problems, and in the end, we managed to retrieve the result using
🏆 Accomplishments that we're proud of
- We built the sort of good-looking app in the span of the span of the hackathon, from scratch.
- We managed to interface the ML model with the app and finally managed to complete it.
📖 What we learned
- Image Classification using Convoluted Neural Networks
- Implementing ML on mobile devices using tflite
- Interfacing tflite and flutter using tflite_flutter
🚀 What's next for the app
Pitch the app to angel investors and become billionaires, support such other great innovations in the future.
Built With
- flutter
- keras
- tensorflow
Log in or sign up for Devpost to join the conversation.