insect-detect-csv
All format examples · Machine-readable schema · Convert outputs to COCO
Frame-normalized corners, detector confidence, track identity and camera metadata. Exposure is milliseconds. Timestamp has no timezone. Use session and filename together for image identity.
Example
Illustrative values, not predictions from a model. Unless stated otherwise, the image is 200 pixels wide and 100 pixels high.
device_id,session_id,timestamp,label,confidence,track_id,track_status,x_min,y_min,x_max,y_max,lens_position,iso_sensitivity,exposure_time,filename
trap-1,1,2026-09-30T12:00:00,insect,0.8,7,TRACKED,0.05,0.2,0.25,0.6,130,100,10,frame.jpg
Convert to a standard format
This format is documented, but a conversion adapter is not yet registered.
General conversion example and required context
How to interpret it
Scope: Camera-trap session metadata CSV for active DepthAI tracklets. Full hardware inference remains unverified because no OAK device is attached.
Coordinates: x_min,y_min,x_max,y_max are frame-normalized corners after deletterboxing, rounded to four decimals; multiply x by original frame width and y by original frame height for pixels.
Confidence: Detector confidence in [0,1], rounded to two decimals; label is the configured display name rather than a numeric class ID.
Tracking: Only NEW and TRACKED tracklets are written; LOST and REMOVED have no CSV rows. track_id is a tracker identifier, not a class index.
Timestamp: timestamp is datetime.now().isoformat(): local time without a timezone suffix, so it is declared string, not the stricter schema ISO type.
Representation: One CSV table decoded using the header as field names. Numeric cells may be safely cast to the declared types; the author writes CSV, not JSON objects.
Evidence
Source inspection at the pinned revisions; no full inference observation claimed.