WearifAI | AI Fashion Cloth Changer

Open-Sourced

Github Repository

Pretrained Models

stabilityai/sd-vae-ft-mse

Raw Dataset

2400+ images

Training Parameter

512x512

Pipeline

OpenPose-based pose detection

The Challenges

Traditional AI models for text-to-image generation often lack the ability to accurately customize clothing items in a photorealistic manner. This limitation is particularly pronounced in applications where users require personalized imagery (e.g., showing a specific type of hoodie or jacket on a person) with high visual fidelity and consistent style. Additionally, managing and processing diverse datasets, along with ensuring compatibility across evolving deep learning libraries, adds further complexity.

The Solution

The solution involves a multi-stage approach that customizes and optimizes the process of generating photorealistic images with personalized clothing. By leveraging DreamBooth for fine-tuning, the process begins with cleaning and categorizing the dataset to ensure high-quality training data. Once trained, the model integrates advanced inference methods by combining inpainting with ControlNet’s pose detection (using OpenPose) to dynamically adapt to human postures. 

  • Customized Fine-Tuning
    Utilize DreamBooth to adapt a pretrained diffusion model to generate images with custom clothing. Train with specific instance and class prompts for accuracy and reproducibility.
  • Optimized Training Environment:
    Apply 8-bit Adam optimizer and manage CUDA dependencies to efficiently use limited GPU memory.
  • Advanced Inference Pipeline:
    Integrate inpainting with ControlNet using OpenPose for dynamic human pose detection and editing.

all Projects