Dataset Ninja LogoDataset Ninja:

Texture Atlas Extractor [hot] Jun 2026

1618519721209
Taggeneral, benchmark
Taskobject detection
Release YearMade in 2013
Licenseunknown

Texture Atlas Extractor [hot] Jun 2026

– train a detector to segment sprite boundaries. Rare due to high variability.

| Feature | Description | |---------|-------------| | | Parse several atlas images and metadata files together. | | Format conversion | Extract to PNG, WebP, TGA, etc. | | DDS decompression | Handle block‑compressed atlases (BC1‑BC7) via libraries like dds‑py or nv_dds . | | Packing data repair | Re‑create metadata by hashing extracted sprites. | | Batch mode | Process entire folders. | | GUI preview | Show atlas with draggable rectangles for manual extraction. | texture atlas extractor

: You lost the original individual source files but still have the final atlas. – train a detector to segment sprite boundaries

You have atlas.png but no .json or .plist . Solution: You need a visual extractor. Tools like SpriteCutter or ImageMagick with grid detection can attempt to slice based on transparent guides. If the sprites are uniformly packed in a grid (e.g., 8x8 icons), you can manually set rows and columns. | | Format conversion | Extract to PNG, WebP, TGA, etc

A (or unpacker) is a developer tool designed to reverse the "packing" process, slicing a single large image—containing multiple smaller sprites or textures—back into individual files.

For each entry in metadata_file.json: 1. Read "filename" (e.g., "sword_iron.png") 2. Read "frame": "x": 100, "y": 50, "w": 32, "h": 32 3. Create a new blank bitmap (width: 32, height: 32) 4. For y_offset from 0 to 31: For x_offset from 0 to 31: source_pixel = atlas_pixel[x + x_offset, y + y_offset] new_pixel[x_offset, y_offset] = source_pixel 5. Save new bitmap as "sword_iron.png"