Project Overview

CultivART is an open-source mobile/web application that allows FarmBot users to capture their plants and create AI-generated images of them. The project is a university final year project that I lead in a team of 4 with an industry client.

Goal

The goal of the project was to create a mobile + web application that allows FarmBot users to capture their plants and create AI-generated images of them. The project was to be used as an art installation at an art gallery in Sydney, to be interacted by the public via an iPad to showcase the capabilities of the AI.

For our case, we had to develop a system that could reliably generate artworks based on the plant data (ranging between atmospheric data and an image capture of the plant via a FarmBot). We also chose to rework the front-end of the website as the provided website was not mobile-friendly and did not have a good user experience.

Key Features

• FarmBot Integration: The system integrates with the FarmBot API to capture the plant data and generate the artwork. Has access to moving the arm location, turn on the lights, and take a photo of the plant.
• AI-Generated Artwork: The system uses a set of self-trained LoRA models to generate the artwork based on the plant data. This system was built using ComfyUI, allowing artists to fine-tune temperatures and other parameters to generate the artwork.
• AI Queue Management: The system was designed with scale in mind, allowing multiple GPU servers to be used to generate the artwork. This allows the system to generate multiple artworks at once reducing the overall time to generate an artwork.
• Admin Dashboard: The system has a custom-built admin dashboard that allows the user to manage the system, users and manage artworks.