Skip to main content

Preprocessing Overview

The preprocessing module provides comprehensive utilities for preparing images for virtual try-on applications.

Features​

  • Garment Segmentation: Segment garments into categories (upper, lower, dress)
  • Garment Extraction: Extract and preprocess garments
  • Human Segmentation: Isolate human subjects from images
  • Pose Estimation: Extract pose keypoints using OpenPose
  • Image Captioning: Generate captions from fashion images

Module Structure​

tryon/preprocessing/
├── preprocess_garment.py      # Batch garment processing
├── extract_garment_new.py      # Single image garment extraction
├── preprocess_human.py         # Human segmentation
├── captioning/                 # Image captioning
├── u2net/                      # U2Net segmentation models
└── utils.py                    # Utility functions

Quick Reference​

Garment Segmentation​

from tryon.preprocessing import segment_garment

segment_garment(
    inputs_dir="path/to/input",
    outputs_dir="path/to/output",
    cls="upper"  # "upper", "lower", "dress", or "all"
)

Garment Extraction​

from tryon.preprocessing import extract_garment

extract_garment(
    inputs_dir="path/to/input",
    outputs_dir="path/to/output",
    cls="upper",
    resize_to_width=400  # Optional
)

Human Segmentation​

from tryon.preprocessing import segment_human

segment_human(
    image_path="path/to/image.jpg",
    output_dir="path/to/output"
)

Supported Garment Categories​

  • upper: Upper body garments (tops, shirts, jackets)
  • lower: Lower body garments (pants, skirts)
  • dress: Full-body dresses
  • all: Automatically detect and process all categories

Next Steps​