The problem DeepEye solves

With the rapid advancements in the field of AI, it becomes very difficult to judge whether what you're surfing on the internet is real or not. Could it be a deep fake? Is the image/video/audio displayed edited? The problem is huge.

A deep fake detector, in which the user can input an image file in our model, and the model, by using classification analysis, can determine whether the input file is deep fake or not.

Challenges we ran into

The problems faced while designing the project were huge. With the intent to design a consumer-based UI application, it became difficult to tackle once it was found that the architecture of the newly released version of TensorFlow was not compatible. Converting it to a JSON required to use of a tensorflowjs converter, but the supported versions were outdated. We tried to switch our method of implementation to create an open-source API for developers, monetizing it so that we could charge for bigger files. But, the model could not be consumed.

The hackathon came with a bunch of turnarounds. It was tiresome, hectic, and in a way, full of emotions and hard work. The amazing work of our team leader, Mr. Akshay Prabhat Mishra, truly proved to be excellent.