Downloads · 30 days
0
JcProg/pedsense-yolo26m-detector-jaad-aug-10e
pedsense-yolo26m-detector-jaad-aug-10e is a object detection model from JcProg. Use it when you need objects located in an image. The card lists the license as agpl-3.0.
A single-class pedestrian detector fine-tuned on the JAAD (Joint Attention in Autonomous Driving) dataset using the PedSense-AI framework. Trained with custom data augmentation as a comparison against the baseline 10-…
Downloads · 30 days
0
Access
Public
Updated Mar 4, 2026
Repo size
88.7 MB
Likes
0
Public
Click a slice to open those files.
.pt88 MB · 99%
From the Hugging Face model README
A single-class pedestrian detector fine-tuned on the JAAD (Joint Attention in Autonomous Driving) dataset using the PedSense-AI framework. Trained with custom data augmentation as a comparison against the baseline 10-epoch model.
| Property | Value |
|---|---|
| Architecture | YOLO26m |
| Task | Object detection (1 class: pedestrian) |
| Dataset | JAAD |
| Epochs | 10 |
| Image Size | 640x640 |
| Batch Size | 16 |
| Train/Val Split | 80/20 (video-level) |
| Parameter | Baseline | This model |
|---|---|---|
degrees | 0.0 | 5.0 |
scale | 0.5 | 0.7 |
mixup | 0.0 | 0.1 |
mosaic | 1.0 | 1.0 |
fliplr | 0.5 | 0.5 |
| Metric | Value |
|---|---|
| Precision | 0.859 |
| Recall | 0.713 |
| mAP@50 | 0.809 |
| mAP@50-95 | 0.494 |
from ultralytics import YOLO
model = YOLO("weights/best.pt")
results = model("image.jpg")
The model was trained using the PedSense-AI framework:
uv run pedsense preprocess frames
uv run pedsense train -m yolo-detector -n detector_aug -e 10 -b 16 \
--yolo-variant yolo26m \
--aug-degrees 5 --aug-scale 0.7 --aug-mixup 0.1
AGPL-3.0