quality-settings

Optimize Quality Settings

Recommend processing parameters tuned to the user's dataset and hardware.

Information Needed

Ask the user (or read from the project):

• Photo count: Number of images in the dataset
• Available RAM: System memory in GB
• Priority: Quality vs speed

Guidelines by Dataset Size

Setting	<100 photos	100–500	500–2000	2000+
Match downscale	1 (High)	1 (High)	2 (Medium)	2–4
Keypoint limit	60000	40000	40000	40000
Depth map quality	2 (High)	4 (Medium)	4 (Medium)	8 (Low)
Depth filter	mild	mild	moderate	moderate
Face count	high	high	medium	medium/custom

RAM Considerations

• <16 GB: Use downscale 4+ for depth maps, medium face count
• 16–32 GB: Use downscale 2–4, high face count up to ~500 photos
• 32–64 GB: Use downscale 1–2, high face count up to ~1000 photos
• 64+ GB: Can use ultra quality for smaller datasets

GPU/CPU Rule

• set_gpu_config(cpu_enable=True) — BEFORE alignment (match_photos, align_cameras)
• set_gpu_config(cpu_enable=False) — BEFORE everything else (depth maps, meshing, texturing, DEM, ortho)

CPU slows GPU operations. It is ONLY beneficial for alignment.

Output

Read the chunk info, then provide specific parameter values for each processing step as ready-to-use tool calls.