Artist Asylum

Generate your art with just a text using Artificial Intelligence which is based on Generative Adversarial Networks. Use Artificial Intelligence to caption your image, which is based on LSTM.

Comment

Inspiration

  • With the rise of AI, it becomes easier for artists to express themselves without having to go through the process of creating an artwork.
  • Artificial intelligence has the ability to consistently learn from past experiences and adapt to changes, making it the perfect candidate for eventually generating meaningful captions.

What it does

  • A single line of text can be used to generate art. The model will generate the art based on the text provided. AI Art generator
  • The AI-based caption generator can generate captions for your images within seconds. AI caption generator AI caption generator

How we built it

  • The AI art generator was created using Generative adversarial network and Imagenet 16384.
  • The AI caption generator was developed using LSTM and was trained on the flickr dataset.
  • HTML, CSS, and Bootstrap were used to create the user interface.
  • We used Django as our backend to integrate our Deep learning model with the UI.

Challenges we ran into

  • Deep learning model training required a significant amount of time and computation power.
  • It was difficult to integrate our deep learning model with the user interface.
  • Improving the model's accuracy was an aesthetic taste.

Accomplishments that we're proud of

  • We were able to create an amazing model that can generate art and captions for images. 
  • Possibly developing a web application and incorporating this model was incredible.

What we learned

We discovered a lot about Deep Learning models and their applications. We got some knowledge on MLOps. We most likely learned more about GANs and LSTM.

What's next for Artist Asylum

  • Should try to reduce the process time for each request.
  • Improve the AI art generator with more options which can make the art generation flexible.
  • Working on to improve the accuracy of the models.

Built With

Share this project:

Updates