AI-COPS

Table of Contents

• Dependencies
• Setup
• Project Description
• Mugshot Generator
• Drawing to Hyperrealistic Portrait
• Knowledge Graph Generator

Dependencies

You'll need a OpenAI account, and a Replicate (replicate.com) account.

Setup

pip install -r req.txt

Edit the .env file to include your API keys.

streamlit run Mugshot-Generator.py

Project Description

This project leverages generative AI models to create mugshots and hyper-realistic portraits from textual descriptions and drawings, while also generating knowledge graphs from prompts, aiding police and investigative agencies in their efforts.

Mugshot Generator

• Utilizes the Stable Diffusion XL (SDXL) model from Stability AI to generate mugshots of suspects or perpetrators based on a textual description.
• Allows users to provide a detailed prompt, specifying the desired characteristics of the subject's face, such as gender, age, ethnicity, and other distinguishing features.
• The generated mugshots are highly realistic, with sharp focus and ultra-high definition (4K) quality.

Drawing to Hyperrealistic Portrait

• Leverages the power of ControlNet, a technique that allows for better control over the image generation process by providing a reference image or drawing.
• Users can upload a drawing or sketch of a face, and the model will generate a hyper-realistic portrait based on the input.
• The generated portraits are highly detailed and lifelike, capturing the essence of the original drawing while enhancing it with realistic textures, lighting, and depth.

Knowledge Graph Generator

• Utilizes the GPT-3.5-turbo model from OpenAI to extract relationships and entities from a given prompt.
• Based on the provided prompt, the model generates a list of relationships in the format [ENTITY 1, RELATIONSHIP, ENTITY 2].
• The extracted relationships are then visualized as a knowledge graph using the NetworkX library in Python, allowing for easy interpretation and understanding of the underlying connections.