Downloads · 30 days
33
12% of all-time downloads
mot20-yolo26s-pedestrian
Halftom/mot20-yolo26s-pedestrian
mot20-yolo26s-pedestrian is a object detection model from Halftom. Use it when you need objects located in an image. It is set up for ultralytics. The card lists the license as mit.
YOLO26s fine-tuned on the MOT20 benchmark for pedestrian detection in dense crowd scenes. This repository contains the trained best.pt weights for pedestrian detection. Multi-object tracking is performed separately us…
All-time downloads
270
Public
Repo size
20.4 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.pt20.4 MB · 100%
At a glance
- Task
- Object Detection
- Library
- ultralytics
- License
- mit
- Access
- Public
- Created
- Jun 19, 2026
- Updated
- Jun 25, 2026
- SHA
- dbb13e25
Model card
From the Hugging Face model README
🚶 MOT20 YOLO26s Pedestrian Detector
YOLO26s fine-tuned on the MOT20 benchmark for pedestrian detection in dense crowd scenes. This repository contains the trained best.pt weights for pedestrian detection. Multi-object tracking is performed separately using ByteTrack and is available in the accompanying GitHub repository and Hugging Face Space.
Model Details
| Item | Value |
|---|---|
| Architecture | YOLO26s |
| Task | Pedestrian Detection |
| Dataset | MOT20 |
| Classes | Person |
| Training Resolution | 640 × 640 |
| Evaluation Resolution | 1280 × 1280 |
Training Data
Training Sequences
- MOT20-01
- MOT20-03
- MOT20-05
Validation Sequence
- MOT20-02
The validation sequence was completely withheld during training.
Detection Performance
Evaluation was performed on the MOT20-02 validation split using:
imgsz = 1280
conf = 0.25
iou = 0.50
| Metric | Value |
|---|---|
| Precision | 94.57% |
| Recall | 91.30% |
| [email protected] | 92.10% |
| [email protected]:0.95 | 67.49% |
The detector achieves high pedestrian detection accuracy on dense crowd scenes, providing reliable detections suitable for downstream multi-object tracking.
Tracking Reference
When combined with ByteTrack, the complete tracking pipeline achieved the following results on MOT20-02:
| Metric | Value |
|---|---|
| HOTA | 50.0% |
| DetA | 55.7% |
| AssA | 45.1% |
| LocA | 85.0% |
Note: These tracking metrics are reported for the complete YOLO26s + ByteTrack pipeline and are not produced by the detector alone.
Usage
Load the Model
from ultralytics import YOLO
model = YOLO("best.pt")
Run Inference
results = model.predict(
source="image.jpg",
imgsz=1280,
conf=0.25
)
Download from Hugging Face
from huggingface_hub import hf_hub_download
from ultralytics import YOLO
weights = hf_hub_download(
repo_id="Halftom/mot20-yolo26s-pedestrian",
filename="best.pt"
)
model = YOLO(weights)
Related Resources
📂 GitHub Repository
https://github.com/Haftom19/MOT20-Pedestrian-Tracking-YOLO26s-ByteTrack
🚀 Interactive Gradio Demo
https://huggingface.co/spaces/Halftom/mot20-pedestrian-tracking
License
MIT License
- Task
- Object Detection
- Library
- ultralytics
- License
- mit
- Created
- Jun 19, 2026
- Updated
- Jun 25, 2026