ShelfApi

Green lines show detected bay boundaries. Orange lines show top/bottom boundaries. Right side shows the result after perspective correction and blur masking.

Shelf — nearly straight-on

Original + detected boundaries

➔

Corrected + masked

Shelf — slight perspective skew

Original + detected boundaries

➔

Corrected + masked

Shelf — significant perspective distortion

Original + detected boundaries

➔

Corrected + masked

Endcap display (kopstelling) — left/right detection

Original + detected boundaries

➔

Corrected + masked

Checkout display (kassameubel) — top/bottom + left/right detection

Original + detected boundaries

➔

Corrected + masked

Process a shelf photo (returns image):

# Blur masking with perspective correction (default)
curl -X POST -F "image=@photo.jpg" \
  "https://shelfapi.wiggert.nl/api/process-shelf" \
  -o result.jpg

# Crop to detected bay only
curl -X POST -F "image=@photo.jpg" \
  "https://shelfapi.wiggert.nl/api/process-shelf?method=crop" \
  -o cropped.jpg

# Black masking + level shelves
curl -X POST -F "image=@photo.jpg" \
  "https://shelfapi.wiggert.nl/api/process-shelf?method=black&correct_horizontal=true" \
  -o result.jpg

# No perspective correction, PNG output
curl -X POST -F "image=@photo.jpg" \
  "https://shelfapi.wiggert.nl/api/process-shelf?correct_perspective=false&output_format=png" \
  -o result.png

Detect boundaries only (returns JSON):

# CNN detector (fast, ~50ms)
curl -X POST -F "image=@photo.jpg" \
  "https://shelfapi.wiggert.nl/api/detect-meters"

# Example response:
{
  "left_top_pct": 12.3,
  "left_bottom_pct": 14.1,
  "right_top_pct": 87.5,
  "right_bottom_pct": 85.9,
  "left_boundary_pct": 13.2,
  "right_boundary_pct": 86.7,
  "top_boundary_pct": 0.0,
  "bottom_boundary_pct": 100.0,
  "source": "cnn"
}

Dataset statistics:

curl "https://shelfapi.wiggert.nl/api/dataset/stats"

# Returns: {"total_images": 3814, "annotated": 3408, "verified": 880, ...}