I’m building an end-to-end “Car Photo AI Factory” that takes ordinary vehicle images and returns studio-quality shots with flawless color, balanced lighting, and upgraded or fully replaced backgrounds—all while preserving a realistic look. What I need: • A robust deep-learning pipeline that detects the car, masks it cleanly, and applies color-grading and relighting to match showroom standards. • A background-swap module offering a curated library (dealership floors, urban streets, neutral gradients) plus the option to upload custom scenes. • Strict realism controls so the final output never looks synthetic, cartoony, or vintage; everything should feel like a high-end commercial shoot. • A scalable inference API (Python/FastAPI or similar) and a lightweight web dashboard for bulk uploads, side-by-side previews, and batch download. • Training scripts, model weights, and clear documentation so my in-house team can retrain or extend the system later. Key tech I expect to see: PyTorch or TensorFlow, state-of-the-art segmentation (e.g., SAM, Mask R-CNN), HDR relighting or neural re-illumination, and GAN-based background synthesis where needed. Feel free to suggest superior alternatives if they keep results photorealistic. I’ll provide sample datasets and a small benchmark set for acceptance testing. Delivery milestones should cover architecture design, prototype models, refinement rounds, and final deployment on my AWS account. If crafting stunningly realistic car imagery at scale excites you, let’s talk details and timeline.