Logo Avatars Pipeline Repository

Blog Explanation: Medium

Overview

This repository contains the backend of a Stable Diffusion application, specifically designed for generating creative logos. The app utilizes Stable Diffusion version 1.5 integrated with a Control Net Canny model to transform standard logos into more creative versions.

Streamlit Frontend Code Here

Features

• Stable Diffusion 1.5 with Control Net Canny: Leverages the latest in AI to generate unique logo designs.
• FastAPI Implementation: Built using FastAPI for efficient and scalable backend services.
• Base64 Image Encoding: Accepts and processes images encoded in Base64 format, streamlining the image handling process.
• Seamless Inference Process: Conducts inference on the decoded images and returns the creatively enhanced logos in Base64 format.

Usage

• The FastAPI backend accepts a 'BaseSDRequest' object containing all necessary parameters.
• Images should be sent as Base64 encoded strings. These are then decoded at the backend for processing.
• After inference, the final image is re-encoded in Base64 and returned to the user.

Google Colab Notebook

Included in the repository is a Google Colab notebook for the Logo Avatars POC. This interactive notebook allows you to run the entire process and see the results in real-time. Just copy link of any image & run it. Colab Notebook

API Reference

Check Heartbeat

  GET /heartbeat

Parameter	Type	Description
none	none	To check if the server is running

Generate Image

  POST /generate_image

Parameters

Name	Type	Description
prompt	string	Required. The prompt for generating the image.
encoded_control_net_image	string	Required. The encoded control net image in base64 format.
control_type	string	Required. The type of control used in image generation.
height	int	Required. The height of the generated image.
width	int	Required. The width of the generated image.
controlnet_conditioning_scale	float	The scale of control net conditioning. Default is 1.0.
negative_prompt	string	A string specifying negative prompts. Default is "deformed, nsfw, blurr".
base_model	string	The base model used for generation. Default is "digiplay/Juggernaut_final".
num_inference_steps	int	The number of inference steps. Default is 20.
guidance_scale	float	The guidance scale. Default is 0.6.
num_images_per_prompt	int	The number of images generated per prompt. Default is 1.

{
  "prompt": "Colorful, jungle surrounding, trees, natural, detailed, hd, 4k, best quality, extremely detailed",
  "generated_image_encoded": "base64_encoded_generated_image_string"
}

Support Me: