Ablation Experiment Results for UAV Small Object Detection

figshare hosts ablation experiment results from the DPCNet model tested on the HIT-UAV dataset. The data, in an XLS file of 5.5 KB, compares model performance against a YOLO11n baseline, showing a 5.1% [email protected] improvement.

Use Cases

• Compare [email protected] improvements between DPCNet and the YOLO11n baseline for small object detection.
• Analyze the relationship between parameter count reduction and performance gains in UAV imagery.
• Evaluate precision and recall metrics for small, dense, low-light, and occluded targets in ablation studies.

Strengths

• Provides concrete performance metrics, including a 5.1% [email protected] improvement on HIT-UAV.
• Reports a specific model architecture detail: a parameter count reduction of about 45%.
• Data is openly licensed under CC BY 4.0.

Limitations

• The dataset is extremely small at 5.5 KB, offering limited scope for statistical analysis or model training.
• Unknown row and column structure restricts understanding of the data's granularity and analytical potential.

Provenance

Source

figshare, authored by Linfeng Jia.

Collection Method

Results from ablation experiments on the HIT-UAV dataset for the DPCNet object detection model.

Time Range

null

Freshness

null

Geography

null