TamAko783/YOLO26s_RDD_Base

YOLO26s · Base (small, GT-only)

A YOLO26s road-damage detector (4-class CRDDC: D00 longitudinal, D10 transverse, D20 alligator, D40 pothole) on the Unified Road Defect Dataset.

Method: Ground-truth only — no distillation. Larger-capacity supervised baseline.

Metrics — RDD held-out validation (4,509 images, 11,470 boxes)

Class	mAP@50	mAP@50-95	Precision	Recall	F1
D00 Longitudinal	0.630	0.351	0.699	0.564	0.625
D10 Transverse	0.669	0.352	0.710	0.583	0.640
D20 Alligator	0.715	0.396	0.742	0.640	0.687
D40 Pothole	0.733	0.388	0.774	0.649	0.706
Overall	0.687	0.372	0.731	0.609	0.665

Full model comparison (same held-out val)

All five models in this study, evaluated identically (imgsz 640):

Model	Variant	Params	Distillation	mAP@50	mAP@50-95	F1
YOLO26n_RDD_Base	YOLO26n	2.4M	— (GT only)	0.635	0.334	0.621
YOLO26n_RDD_FRDC_Distilled	YOLO26n	2.4M	1 teacher (Co-DETR)	0.640	0.337	0.625
YOLO26n_RDD_FRDC_Distilled_v2	YOLO26n	2.4M	2 teachers (Co-DETR+RTMDet)	0.638	0.337	0.626
➤ YOLO26s_RDD_Base (this model)	YOLO26s	9M	— (GT only)	0.687	0.372	0.665
YOLO26s_RDD_FRDC_Distilled_v2	YOLO26s	9M	2 teachers (Co-DETR+RTMDet)	0.692	0.375	0.672

Reading it:

• Distillation helps — every distilled model beats its GT-only baseline.
• Capacity helps most — the YOLO26s models (+~0.05 mAP@50) clearly outperform the nano ones on this val.
• One vs two teachers is a near-tie at nano size; the two-teacher set's edge is small.
• Cross-domain check (independent RDDC2024-ID, Indonesia, 8,901 imgs): distilled models generalized better than baselines, while the larger GT-only model generalized worse — evidence the distillation's added data improves robustness, not just in-domain fit.

RDD ground truth has known missing annotations, so absolute precision/recall are conservative for all models. The comparison is fair — every model uses the identical held-out val, never trained on.

Usage

from ultralytics import YOLO
model = YOLO("YOLO26s_RDD_Base.pt")
results = model("road.jpg")

Models in this suite

• YOLO26n_RDD_Base · YOLO26n_RDD_FRDC_Distilled (v1) · YOLO26n_RDD_FRDC_Distilled_v2
• YOLO26s_RDD_Base · YOLO26s_RDD_FRDC_Distilled_v2

Datasets

• Base: TamAko783/Unified_Road_Defect_Dataset
• Distillation sets: v1 · v2 (two-teacher)

Credits

• Datasets: RDD-2022 (Arya et al.) · UAV-PDD2023 · RoadDamageVision (Silva Zendron & Leithardt, CC BY 4.0).
• Distillation teachers: Co-DETR Swin-L + RTMDet-x — FRDC (Wang Fangjun et al.), ORDDC'2024 winner.
• Student: YOLO26s, AGPL-3.0 (Ultralytics).