YOLOv8: real-time object detection

YOLOv8 is Ultralytics' real-time detector. It finds objects of COCO's 80 classes and returns a box, a class and a confidence for each. Five sizes, from nano at 7 MB to xlarge at 137 MB. Its weights are AGPL-3.0, which matters if you serve predictions over a network: read Weights and licence before you deploy it.

Quick start

Python
import mozo

model = mozo.get_model("yolov8/nano")
found = model.predict("example.jpg")

for detection in found[:4]:
    print(detection.class_id, detection.class_name, round(detection.confidence, 2), detection.bbox)
Text only
0 person 0.92 [0.0, 113.24, 763.19, 1269.33]
0 person 0.92 [1295.3, 17.81, 1920.0, 1269.44]
63 laptop 0.85 [709.09, 773.48, 1366.4, 1143.08]
41 cup 0.64 [632.86, 589.76, 704.29, 724.06]

Over HTTP, with mozo start running:

Bash
curl -X POST "http://localhost:8000/predict/yolov8/nano?threshold=0.5" -F "file=@example.jpg"

The response is the same detections as a JSON list, with every PixelFlow field on every detection.

Variants

ID Task Weights (torch) Runtimes Licence
yolov8/nano detection 7 MB torch, onnx, coreml AGPL-3.0
yolov8/small detection 23 MB torch, onnx, coreml AGPL-3.0
yolov8/medium detection 52 MB torch, onnx, coreml AGPL-3.0
yolov8/large detection 88 MB torch, onnx, coreml AGPL-3.0
yolov8/xlarge detection 137 MB torch, onnx, coreml AGPL-3.0

runtime="auto", the default, picks CoreML on Apple silicon (mps) and torch on CUDA and CPU. CoreML is by far the fastest way to run YOLOv8 on a Mac. To choose a runtime yourself, load the model through a ModelManager, which takes runtime:

Python
from mozo import ModelManager

model = ModelManager().get_model("yolov8", "nano", runtime="onnx-fp32")   # pip install 'mozo[onnx]'

YOLOv8 has no segmentation variants in mozo. For masks from a YOLO, see YOLO11 or YOLO26.

Inputs and outputs

Python
model.predict(image, threshold=0.5, labels=None)
Parameter Type Default Description
image str, bytes or numpy.ndarray A file path, encoded image bytes, or an HxWx3 RGB array.
threshold float 0.5 Minimum confidence to keep a detection.
labels list[str] None Class names for this call, replacing the checkpoint's own.

Returns a PixelFlow Detections, best first. Each detection has:

Over HTTP the same parameters are query parameters: threshold, and labels as a comma-separated list (labels=hardhat,vest).

How it behaves

Overlaps are suppressed

YOLOv8 predicts many overlapping boxes per object and removes the duplicates with non-maximum suppression, as Ultralytics does. That step runs inside mozo, identically whichever runtime ran the network. YOLO26 is the YOLO that needs no suppression step.

Class ids are contiguous

The YOLO families number COCO's classes 0 to 79, so person is 0. RF-DETR uses COCO's original ids, where person is 1. This matters when you filter by id, for example before handing people to ViTPose:

Python
people = found.filter_by_class_id([0])

Class names come from the checkpoint

The names are read from the weights, so a fine-tuned YOLOv8 brings its own vocabulary. Run from an ONNX or CoreML graph, which carries no names, mozo reads them from a labels file published beside the graph.

Weights and licence

The weights are Ultralytics' ultralytics/assets release v8.4.0, redistributed unmodified. They are AGPL-3.0, or covered by a commercial licence from Ultralytics.

Mozo's own code is Apache-2.0, and so is its YOLOv8 inference: it was written independently against the checkpoint format, and it imports, contains and depends on nothing from ultralytics. The code and the weights are separate works that travel together. But anything you export from the weights inherits their terms, and serving predictions from them over a network places AGPL-3.0 section 13 obligations on you. Complying is the operator's responsibility.

Every variant ships with the licence and a NOTICE naming the release it came from. To run a server that never offers AGPL weights, leave the YOLO families out of MOZO_ENABLE. The details are in mozo/vendors/yolov8_deploy/PROVENANCE.md.