ReviveAI

Bringing the Past into the Present.

Created on 27th October 2024

•

ReviveAI

Bringing the Past into the Present.

The problem ReviveAI solves

ReviveAI is an advanced AI-powered platform that transforms old black-and-white photos into vibrant, high-definition images. Using state-of-the-art technology, it not only colorizes but also enhances the quality of historical photos, revealing details, textures, and colors as if they were captured yesterday. With ReviveAI, anyone can reconnect with the past and preserve family history in stunning, true-to-life visuals.

Challenges we ran into

Scaling and training the models due to computational constraints: We used pre-trained models in the cases where we were not able to train the model by ourself. We trained Enhancer (ESRGAN) on CUDA on our dataset acquired from Kaggle while Colorization, Stylize and Stable Diffusion models were already pre-trained.
API integration due to differences in input and output schema: Management of various nodes of different types such as input, output, model and differentiating them further based on their characteristics.
Finding models that can be used for our specific use case: Extensive search with hit and trial method to choose pretrained-models that can be finetuned further.
Integration of webcam: Frame capture one moment at a time. We might work with real-time processing if the opportunity arrives.
Video processing: Integrated videos as part of input by frame-to-frame slicing and processing each frame individually. We tried parallel processing of each frame but were unable to successfully come up with any results in given timeframe.

Technologies used

Node.js

Flask

TensorFlow

PyTorch

Postman

Python

TypeScript

MongoDB

Generative Advarsarial Networks (GANs)

React.js

Discussion

Builders also viewed

See more projects on Devfolio