Pets

/petsdemo

Dogs, cats and birds around the house.

Preparing the model…

Confidence0.25

Head

40.9 mAP

—ms

0 objects

pre: —
infer: —
post: —
total: —

dog
cat
bird

mAP50-95

40.9

Published COCO val2017 score for all 80 classes of the underlying YOLO26n model, not for this filtered subset.

Classes

Filtered from the 80 classes the graph emits.

Input

640px

Letterboxed square input.

Model size

9.9 MB

5.0 MB in fp16, which is what WebGPU clients download.

Sample training images

Nothing to show here yet — this demo runs a COCO-pretrained model rather than one Spotr trained, so there is no generated dataset behind it. Models you create will show the synthetic images they learned from.

Download

One-to-many head
40.9 mAP
output 1×84×8400 · opset 19
fp32 · 9.9 MB fp16 · 5.0 MB
NMS-free head
40.1 mAP
output 1×300×6 · opset 19
fp32 · 9.9 MB fp16 · 5.0 MB

metadata.json (class names, input size, metrics)

COCO-pretrained YOLO26n (Ultralytics 8.4.166, AGPL-3.0), exported to ONNX at 640px and filtered to 3 classes. Not trained by Spotr.