Tiny machine learning empowers climbing inspection robots for real-time multiobject bolt-defect detection

Keyphrases

• Object Model 100%
• Multi-object 100%
• Inspection Robot 100%
• Bolt Defect Detection 100%
• Tiny Machine Learning 100%
• Microcontroller 66%
• F1 Score 66%
• Inference Time 66%
• Climbing Robot 66%
• Random Access Memory 33%
• Balanced Detection 33%
• Effective Method 33%
• Inspection Technology 33%
• Lite 33%
• Detection Method 33%
• Defect Detection 33%
• Complex Structure 33%
• Detection Task 33%
• Low Capacity 33%
• Environmental Conditions 33%
• Performance Analysis 33%
• 32-bit 33%
• 8-bit 33%
• Inspection Method 33%
• Flash Memory 33%
• System Accuracy 33%
• Structural Integrity 33%
• Steel Structures 33%
• Edge Computing 33%
• Time Application 33%
• Feature Pyramid Network 33%
• New Benchmark 33%
• Cost-effective Solution 33%
• Hardware Requirements 33%
• Real-time Inspection 33%
• Resource Demand 33%
• Inspection Efficiency 33%
• MobileNetV2 33%
• YOLOv5 33%
• Floating-point Data 33%
• Defect Detection System 33%
• Averaged System 33%
• Single Shot multibox Detector 33%
• Recall Value 33%
• Integer Data 33%
• Precision Parameters 33%
• Bolt Failure 33%
• Steel Buildings 33%

Engineering

• Robot 100%
• Defect Detection 100%
• Learning System 100%
• Microcontroller 50%
• Data Format 50%
• Climbing Robot 50%
• Random Access Memory 25%
• Inspection Technology 25%
• Steel Structure 25%
• Flash Memory 25%
• Complex Structure 25%
• Inspection Method 25%
• Structural Integrity 25%
• Bit Floating Point 25%
• Detection Task 25%
• Performance Analysis 25%
• Edge Computing 25%
• Effective Solution 25%
• Display Model 25%
• Steel Construction 25%