AI SKY REPLACEMENT CODING

Full AI Sky Replacement System for Server-Side Implementation (Photoshop Quality)"

--- Overview:

I am looking for an experienced computer vision engineer or AI developer to implement a fully functional, high-quality sky replacement system that runs on a Linux server (ideally FastAPI-compatible). The system should deliver results comparable to Photoshop or Luminar AI, with proper sky detection, segmentation, blending, and lighting correction.


--- Deliverables:

1. Sky segmentation model (pretrained or custom trained)


2. Foreground-background separation with edge handling (trees, hair, etc.)


3. Sky blending & tone matching algorithm


4. Batch API endpoint (e.g., FastAPI) to upload photo + sky and get result


5. Post-processing pipeline (blurring, feathering, color correction)


6. Optional: Skin/human area preservation


7. Complete source code, environment setup, and deployment guide

--- Preferred Technologies:

Python (OpenCV, PyTorch/TensorFlow, NumPy, scikit-image)

FastAPI (for API)

Any efficient model for segmentation: U^2-Net, DeepLabV3+, MODNet, SkyAR, etc.

--- Sample Input/Output:

Please include examples where:

The sky overlaps with trees or hair

There is backlighting or exposure differences

Overcast or gray sky needs vibrant replacement

--- Constraints:

Must run efficiently on GPU-enabled server or optimized CPU fallback

Output should be high-resolution
Open-source base preferred (or license-cleared)

No third-party API (must run locally)

--- Timeline:

Delivery within 1 week including revisions.