Table 1.
Comparative Analysis of various Research work on Object Detection.
| Paper Referred | Dataset | Pixel Size and Resolution | Framework/Algorithm | Precision/Recall/Accuracy |
|---|---|---|---|---|
| [23] | SPOT-5 and Google Earth Services | Pixel Size 9000 × 9000 and Resolution 5 m | Proposed method-based sea surface analysis | Precision- 89.22% & Recall- 97.80% |
| [30] | ImageNet LSVRC-2010 | deep convolutional neural network | Precision- 78.1% & Recall- 60.9% | |
| [32] | high-resolution remote sensing (HRRS) | Pixel Size 600 × 600 | CNN | Accuracy- 94.6% |
| [35] | Kaggle Dataset | Pixel Size 768 × 768 | UNet | Accuracy- 82.3% |
| [36] | Airbus Satellite Image Dataset | Pixel Size 768 × 768 | CNN based Deep Learning | Accuracy- 89.7% |
| [37] | RADARSET-2 and Sentinel-1 | Faster R-CNN | Precision- 89.23% & Recall- 89.14% | |
| [38] | SAR Ship Detection Dataset (SSDD) | Knowledge Transfer Network and CNN based detection model | Precision- 98.87% & Recall- 90.67% | |
| [39] | WorldView-2 and -3, GeoEye and Pleiades | Resolution between 0.3 m and 0.5 m | YOLOV2, YOLOV3, D-YOLO and YOLT | Average Precision- 60% for vehicle and 66% for vessel |
| [40] | Google Earth Images | Resolution be-tween 2 m and 0.4 m | Two staged CNN-based ship detection technique |
Accuracy- 88.3% |
| [41] | Google Earth Images | Pixel Size ranges from 900 × 900 to 3600 × 5400 |
Transfer learned Single-shot Multibox Detector (SSD) |
Accuracy- 87.9% |