px-asset-extract

To extract individual assets into transparent PNGs, you can use an automated pipeline that segments images, classifies components like text blocks and icons, and crops them with anti-aliased alpha edges. This process also generates a JSON manifest for downstream processing.

Yes, you can crop image components using pre-computed bounding boxes by passing region coordinates to the extraction pipeline. This allows you to bridge with grounding models and selectively extract assets from specific areas of slides or diagrams.

You can filter extracted image assets by type during segmentation using command-line flags like --types and --exclude-types. This allows you to specify which classes of components to include or exclude from the final transparent PNG outputs and JSON manifest.

Yes, Python computer vision image segmentation works for batch processing diagrams through a provided CLI and Python API. It supports CPU-based segmentation and classification to decompose multiple diagrams into individual transparent assets.

The best way to automate decomposing slides into individual assets is using a programmatic extraction pipeline that handles segmentation, classification, and cropping. It outputs per-asset transparent PNGs alongside a JSON manifest for tracking.

You need a JSON manifest when extracting image components because it records the classification and metadata for each cropped transparent PNG. This output enables downstream processing, visualization generation, and programmatic tracking of the extracted assets.