Contrek: Vectorizing 10 Billion Pixels on a Raspberry Pi 5 with Just 4 GB RAM

Description and use

Contrek enables processing of 10 billion‑pixel raster images on a Raspberry Pi 5 with 4 GB RAM by streaming horizontal stripes and writing polygons to disk, keeping peak memory below 2.42 GB.

Problem

Processing extremely large raster images without loading the entire image into memory.

Outcome

On a Raspberry Pi 5 with 4 GB RAM, Contrek processed a 10 billion‑pixel image in 13 min 52 s, peaking at 2.42 GiB memory, producing a 21.03 GB SVG without any swap.

Workflow

• 1. Install Contrek on a Raspberry Pi 5 with 4 GB RAM, 250 GB NVMe SSD, Ubuntu 24.04 ARM64 Docker container.
• 2. Prepare a 100,000 × 100,000 pixel PNG image (~900 MB compressed).
• 3. Run Contrek with horizontal stripe width of 1,000 rows.
• 4. Contrek extracts contours, reconnects across stripes, and writes polygons to disk progressively.
• 5. Monitor execution time, peak memory, and CPU usage.
• 6. Verify output SVG (~21 GB) and confirm no swap usage.

Evidence

Contrek can process an image in horizontal stripes, reconnecting contours across stripe boundaries and progressively writing completed polygons to disk.

Source

The input was a 100,000 × 100,000 pixel PNG image, containing 10 billion pixels. The compressed PNG file is about 900 MB.

Source

Peak process memory 2.42 GiB

Source

Output SVG 21.03 GB

Source

The process completed successfully, with zero swaps reported by GNU time.

Source