CuttingLabs ✂️

CuttingLabs is a high-performance image segmentation application that leverages a hybrid AI approach, utilizing Google Gemini API for cloud-based precision and RMBG-2.0 for fast, local background removal. Built with Next.js and Django, it features a professional comparison UI with perfect 1:1 alignment, real-time edge refinement, and intelligent hardware-accelerated fallback to ensure a seamless and efficient cutout workflow.

Key Features

• Hybrid AI Segmentation: Seamlessly switch between Gemini (Cloud) and RMBG-2.0 (Local).
• Pro Comparison Slider: Precision 1:1 alignment for accurate before/after viewing.
• Hardware Acceleration: Automatic detection of CUDA, MPS, or CPU for local processing.
• Edge Refinement: Granular control over Threshold, Feather, and Padding.
• Fast Fallback: Intelligent automatic switching to local model based on API quota or timeout.

Tech Stack

• Frontend: Next.js 14, Tailwind CSS, Lucide Icons.
• Backend: Django, Django REST Framework, Celery, Redis.
• AI Models: Google Gemini Pro Vision, RMBG-2.0 (ONNX).
• Processing: Pillow, NumPy, ONNX Runtime.

Getting Started

Prerequisites

• Python 3.10+
• Node.js 18+
• Docker (optional)

Setup

1. Clone the repository:

   git clone https://github.com/awpetrik/CuttingLabs.git

2. Setup Backend:
   • Create .env from .env.example and add your GEMINI_API_KEY.
   • Install dependencies: pip install -r requirements.txt.
   • Run: python manage.py runserver.
3. Setup Frontend:
   • Install dependencies: npm install.
   • Run: npm run dev.

License

MIT