anyforge/yolo26-human-det

YOLO26 Series Human Detection

Model Description

A human detection model trained on a custom-built dataset using the YOLO26 series.

Model Performance

Dataset

The dataset is aggregated from multiple sources, featuring images of various sizes and scenarios. Key statistics include:

• Total Images: 101,140
• Total Bounding Boxes: 875,283
• Diverse Scenario Coverage
• Sources: Public online datasets and custom-collected data.

Usage Scope

This model detects human positions within input images.

How to Use

• Inference: Input an image; the model returns bounding box coordinates for all detected humans. Supports multi-person detection.

Target Scenarios

• Human detection capabilities applicable to complex and diverse real-world scenarios.

Code Example

### Inference based on Ultralytics
### Install dependencies: pip install modelscope ultralytics

from pathlib import Path
from ultralytics import YOLO

modelfile = Path('/path/to/human-det-yolo26n.onnx')
model = YOLO(modelfile)
imgpath = '/path/to/you/image'
res = model.predict(imgpath)
for line in res:
    print(line.boxes)

Training Process

• Trained for 100 epochs using Ultralytics YOLOv26-v8.4.95.

Evaluation Results

Objective metrics on the validation set:

Model	Class	Validation Images	P
yolov26n	human	5,677	0.93