Animefy

• 

  A splash screen with an animated icon animation. Every time user opens up the app he/she will see a random anime icon.

• 

  Another animated icon splash screen.

• 

  If no image has been selected, app will prompt a message to give it one.

• 

  The image can be selected from the gallery or it can be taken from the camera.

• 

  A pac-man animation will be shown until the tensorflow model gives us the output.

• 

  Then the app will show the matching anime character with respective confidence in percentage

• 

  If the user wants to know more about the anime character he/she can click the bottom button and a detailed info will be shown to the user.

Inspiration

Being weebs and geeks, we always wanted to build something related to anime. providing us the platform to do it, we came up with the Idea to build Animefy. The theme of the Hackathon was perfect for us to express our love for anime.

What it does

Animefy is an Anime Character look-alike app to discover the anime character that you resemble the most. All you have to do is upload your photo and ta-da here appears your anime doppelganger.

How we built it

The whole project was divided into 2 parts building the Face recognition model and building the Application UI. As a group of 4 members we assigned each pair to each part.

• For making the Face detection and recognition part.
• For building the Application we used, the Flutter framework.
• For building ML model we use python language. More specifically we used Keras and transfer learning to build the model with learning rate of 0.001 . The no. of epochs used were 10 and batch size used was 16.

Challenges we ran into

• The main challenge was that the group members had interests in different fields so we found it hard to help each other when facing some problems.
• Integrating TensorFlow model with the Flutter application.
• Initially the accuracy of the model was pretty low. So we had to come up with new ways to increase the accuracy by changing the learning rate of the model and increasing the size of our training dataset.
• We build the app with the aim that it is compatible on as many devices it is possible.

Accomplishments that we're proud of

We finally completed our project despite having many challenges.

What we learned

Challenges we faced helped us learning many new things.

• We learned how to implement a Face recognition model using TensorFlow.
• We learned how to integrate a TensorFlow model with the Flutter application.

What's next for Animefy

We will include as many anime characters possible in the training database of our model. So that the results are more accurate.

Built With

• dart
• flutter
• jikan
• machine-learning
• python
• tflite