Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Benchmarking CNN and Transformer-Based Object Detectors for UAV Solar Panel Inspection

Ashen Rodrigo, Isuru Munasinghe, Pubudu Sanjeewani, Asanka Perera

TL;DR

This work presents a comprehensive benchmark of convolutional and transformer-based object detectors on UAV-captured RGB imagery of solar panels and introduces a class-targeted augmentation strategy applied exclusively to the training split to mitigate imbalance without compromising evaluation integrity.

Abstract

Timely and accurate detection of defects and contaminants in solar panels is critical for maintaining the efficiency and reliability of photovoltaic (PV) systems. While recent studies have applied deep learning to PV inspection, fair benchmarking across detector architectures and unbiased handling of class imbalance remain limited. This work presents a comprehensive benchmark of convolutional and transformer-based object detectors on UAV-captured RGB imagery of solar panels. It introduces a class-targeted augmentation strategy applied exclusively to the training split to mitigate imbalance without compromising evaluation integrity. Faster R-CNN with ResNet50 and MobileNetV3 backbones, RetinaNet with ResNet50, YOLOv5, YOLOv8, and Swin Transformer backbones integrated with Faster R-CNN (Tiny, Small, and Base variants) are evaluated. Performance is assessed using mean Average Precision (mAP) across multiple IoU thresholds, precision, recall, F1 score, and inference throughput to enable accuracy-throughput tradeoff analysis relevant to UAV deployment. Experimental results show that Faster R-CNN with a ResNet50 backbone achieves the highest localization accuracy, with mAP@0.5 of 0.893 and mAP@0.5:0.95 of 0.759, whereas the MobileNetV3 variant provides the best overall reliability balance, achieving recall of 0.745, F1-score of 0.809, and accuracy of 0.679 on the test set. The dataset and code will be released upon acceptance of the paper.

Benchmarking CNN and Transformer-Based Object Detectors for UAV Solar Panel Inspection

TL;DR

This work presents a comprehensive benchmark of convolutional and transformer-based object detectors on UAV-captured RGB imagery of solar panels and introduces a class-targeted augmentation strategy applied exclusively to the training split to mitigate imbalance without compromising evaluation integrity.

Abstract

Timely and accurate detection of defects and contaminants in solar panels is critical for maintaining the efficiency and reliability of photovoltaic (PV) systems. While recent studies have applied deep learning to PV inspection, fair benchmarking across detector architectures and unbiased handling of class imbalance remain limited. This work presents a comprehensive benchmark of convolutional and transformer-based object detectors on UAV-captured RGB imagery of solar panels. It introduces a class-targeted augmentation strategy applied exclusively to the training split to mitigate imbalance without compromising evaluation integrity. Faster R-CNN with ResNet50 and MobileNetV3 backbones, RetinaNet with ResNet50, YOLOv5, YOLOv8, and Swin Transformer backbones integrated with Faster R-CNN (Tiny, Small, and Base variants) are evaluated. Performance is assessed using mean Average Precision (mAP) across multiple IoU thresholds, precision, recall, F1 score, and inference throughput to enable accuracy-throughput tradeoff analysis relevant to UAV deployment. Experimental results show that Faster R-CNN with a ResNet50 backbone achieves the highest localization accuracy, with mAP@0.5 of 0.893 and mAP@0.5:0.95 of 0.759, whereas the MobileNetV3 variant provides the best overall reliability balance, achieving recall of 0.745, F1-score of 0.809, and accuracy of 0.679 on the test set. The dataset and code will be released upon acceptance of the paper.

Paper Structure

This paper contains 14 sections, 8 equations, 5 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Methodology
  4. Datasets
  5. Base Dataset and Class Definitions
  6. Dataset Augmentation Strategy
  7. Dataset Statistics, Annotation Format, and Evaluation
  8. Training Setup
  9. Experiments and Results
  10. Training Configuration and Hyperparameter Settings
  11. Overall Quantitative Performance Evaluation
  12. Confusion Matrix Analysis
  13. Visual Prediction Analysis and Training Convergence
  14. Conclusions and Future Work

Figures (5)

  • Figure 1: UAV-assisted targeted cleaning for solar panels.
  • Figure 2: Representative training samples with ground truth (GT) annotations: (A) Bird droppings, (B) Clean panels, (C) Dust accumulation, (D) Electrical defects, and (E) Physical defects.
  • Figure 3: Confusion matrices of the best-performing three models evaluated on the test dataset.
  • Figure 4: Detection results on solar PV modules with predicted bounding boxes and confidence scores: (A) Bird droppings, (B) Clean panels, (C) Dust accumulation, (D) Electrical defects, and (E) Physical defects.
  • Figure 5: Training and validation loss curves for all evaluated models decrease steadily with minimal divergence, indicating stable learning and generalization to unseen test data.