Vision Transformer Segmentation & Depth-Based Lens Blur

ML Engineer

Selfie segmentation and realistic lens-blur simulation using pretrained segmentation and monocular depth-estimation models.

Highlights

Used a Hugging Face segmentation model to isolate a person from a complex background in 512×512 selfies.
Applied both fixed Gaussian background blur and depth-based variable blur where blur intensity scales with estimated depth.
Packaged the pipeline into a reproducible Colab notebook and optional Hugging Face Space UI for interactive experimentation.