Try this model on one of your photos to see its result here.

Segmentation model

Outlines every object of 80 everyday kinds with a pixel-precise mask. The model behind /segment.

Tasks: Instance segmentation
Outputs: pixel masks (polygons), bounding boxes, class names, confidence scores
Classes (80): The 80 COCO classes. Class names in results are always in English.
personbicyclecarmotorcycleairplanebustraintruckboattraffic lightfire hydrantstop signparking meterbenchbirdcatdoghorsesheepcowelephantbearzebragiraffe

Specifications

Training data: COCO 2017 (118,000 training photos; annotations CC BY 4.0)
Accuracy: Masks
mAP50-95
0.603
mAP50
0.828
COCO val2017, all 5,000 photos, standard COCO metrics, measured by Comexp on the model as served · measured Nov 10, 2026
Masks
mAP50
0.862
LVIS v1 validation set, standard LVIS metrics, measured by Comexp on the model as served · measured Nov 10, 2026
Image requirements: JPG, PNG or WebP, up to 10 MB, any resolution.
Video requirements: Photos only.
Input / output: In: one image, multipart/form-data field file. Out: JSON detections, each with class, confidence, bbox in pixels and polygon.
Licence: Use through the Comexp API under our Terms of service.
Version: Updated Nov 6, 2026
Hardware: Runs in the Comexp cloud: nothing to install, any device that can send an HTTP request. On your own servers or devices: on request.
API: curl -X POST https://ml.comexp.net/api/segment \ -H "Authorization: ApiKey <YOUR_API_KEY>" \ -F "file=@image.jpg"
Each call costs one detection from your plan, as on /detect. Your API key is in your account settings.