Getting the Best Results from Your Camera
Lighting, angle, distance, and framing tips that improve every pipeline's accuracy.
Why camera setup matters
Sproutcast's computer-vision pipelines work directly on pixel data. The segmentation pipeline (P3b.0) uses HSV colour ranges to separate plant from background â which means the colour and brightness of your background and light source directly affect accuracy.
Lighting
Best: Diffused natural light from a north or east-facing window. Consistent and free.
Good: A 5000â6500 K LED panel held to the side of the plant. Daylight colour temperature means the chlorophyll index pipeline isn't fooled by warm light.
Avoid:
- Direct sunlight (creates harsh shadows that confuse edge detection)
- Mixed light sources (e.g. a warm lamp and a cool window at the same time â shifts the white balance unpredictably)
- Backlit shots where the plant is silhouetted
Background
A plain, contrasting background dramatically improves segmentation. A sheet of white or grey cardboard behind a pot plant takes the F1 score from ~0.72 to ~0.94 in our internal tests with *Dahlia pinnata*.
Dark soil in a pot against a white background is ideal â the pipeline distinguishes foliage (greens and yellows) from soil (browns) and background (white/grey) with high confidence.
Distance and framing
- Fill roughly 50â80% of the frame with the plant. Too small and you lose resolution on leaf edges. Too large and you clip leaves â the area metric will be wrong.
- Consistent framing session-to-session matters more than perfect framing. If you keep the camera in the same spot, the growth-rate pipeline (P6.0) can track area changes reliably even if your initial framing wasn't ideal.
Using a fixed mount
A cheap phone tripod or a small articulating arm (the kind used for Ring lights) clamped to a shelf gives you consistent framing across captures. This is especially useful for timelapse mode.