PipeVision AI — Sewer Pipe Defect Detection

Overview

Real-time pipeline inspection analytics

247

Total Inspections

+12 this week

89

Defects Found

36.0% defect rate

94.7%

Detection Accuracy

+1.3% from baseline

1.2s

Avg Processing Time

GPU accelerated

Recent Detections

INS-2026-247

2026-03-22 09:14

Critical

INS-2026-246

2026-03-22 08:47

Major

INS-2026-245

2026-03-21 15:32

Minor

INS-2026-244

2026-03-21 11:05

Critical

INS-2026-243

2026-03-20 16:22

Pass

INS-2026-242

2026-03-20 09:48

Major

Defect Distribution

CR

78%

31

RI

52%

21

DE

40%

16

BP

25%

10

DJ

18%

7

SD

8%

3

IN

3%

1

Legend

CR=Crack RI=Root Intrusion DE=Deposit BP=Broken Pipe DJ=Displaced Joint SD=Surface Damage IN=Infiltration

Defect Detection

Upload a CCTV inspection image to run CNN-based defect analysis

Drop pipe inspection image here or click to browse

Drag & drop or click to select a file
CCTV footage · Lateral camera · Push-rod inspection

JPG PNG BMP TIFF MP4 AVI Max 10MB

Detection Model Ensemble

Image Guidelines

Minimum resolution 640×480px for accurate detection

Ensure adequate lighting; overexposed areas reduce accuracy

Frontal or axial pipe view preferred over angled shots

Avoid heavily motion-blurred or compressed images

Supported Defects

CRCrack RIRoot Intrusion DEDeposit BPBroken Pipe DJDisplaced Joint SDSurface Damage INInfiltration

pipe_section_CCTV_247.jpg 640 × 480 · 1.8 MB

Crack 94.2%

Crack 81.5%

Root Intrusion 87.1%

Deposit 76.8%

Running inference...

YOLOv8-m · GPU · CUDA 12.1

Detection Results

1.24s

Crack

Code: CR · Instance #1

Critical

94.2%

Longitudinal crack · Est. length 180mm · Structural risk

Crack

Code: CR · Instance #2

Major

81.5%

Diagonal hairline crack · Est. length 95mm

Root Intrusion

Code: RI · Instance #1

Major

87.1%

Fine root mass · Cross-sectional blockage 30%

Deposit / Debris

Code: DE · Instance #1

Minor

76.8%

Sediment build-up at invert · Depth ~15mm

Pipe Condition Score

C

Grade C — Requires Attention

Multiple structural defects detected. Recommend maintenance inspection within 3 months. Priority: HIGH.

Inspection History

Browse, filter and export past detection results

to

Date / Time	Inspection ID	Defects	Severity	Model	Conf.
2026-03-22 09:14:32	INS-2026-247	CRRIDE	Critical	YOLOv8-m	94.2%
2026-03-22 08:47:15	INS-2026-246	RIDE	Major	ResNet50	91.7%
2026-03-21 15:32:08	INS-2026-245	DE	Minor	EfficientNet-B4	88.4%
2026-03-21 11:05:44	INS-2026-244	CRBPDJ	Critical	YOLOv8-m	96.1%
2026-03-20 16:22:19	INS-2026-243	None	Pass	YOLOv8-m	99.3%
2026-03-20 09:48:57	INS-2026-242	CRIN	Major	ResNet50	82.6%

Showing 1–6 of 89 results

1

2

3

...

15

Model Information

YOLO-based object detection for sewer pipe defect analysis

Object Detection

YOLOv8-m

Real-time single-stage object detector with anchor-free architecture. Supports bounding box localization and multi-class defect detection in a single forward pass.

Parameters25.9M

Input Size640 × 640

mAP@0.591.4%

Inference8.2ms / GPU

FrameworkUltralytics

TaskDetection + Loc.

Training & Validation Loss

Training Loss

Validation Loss

Epochs →

Training & Validation Accuracy

Training Accuracy

Validation Accuracy

Epochs →

Dataset Information

Source CCTV Sewer Dataset (KITTI-Pipe)

Total Images 4,820

Annotated 4,820 (100%)

Annotation Tool LabelImg / CVAT

Format YOLO / COCO JSON

Augmentation Albumentations

Train / Val / Test Split

Train 70% (3,374) Val 15% (723) Test 15% (723)

Class Distribution

CR

1,250

RI

880

DE

710

BP

515

DJ

385

SD

225

IN

130

Data Augmentation Applied

Random Flip Rotation ±30° Brightness CLAHE GaussNoise Mosaic