Reliable Detection of Minute Targets in High-Resolution Aerial Imagery across Temporal Shifts

TL;DR

This paper presents a transfer learning approach using Faster R-CNN for detecting tiny rice seedlings in high-resolution aerial images, demonstrating robustness across different temporal conditions in UAV datasets.

Contribution

It introduces a curated UAV dataset for rice seedling detection and evaluates the generalization of transfer learning-based detection models across temporal shifts.

Findings

Transfer learning accelerates model convergence in agricultural detection tasks.

The model maintains consistent performance across different temporal UAV datasets.

Curated dataset enables robust evaluation of detection methods in variable conditions.

Abstract

Efficient crop detection via Unmanned Aerial Vehicles is critical for scaling precision agriculture, yet it remains challenging due to the small scale of targets and environmental variability. This paper addresses the detection of rice seedlings in paddy fields by leveraging a Faster R-CNN architecture initialized via transfer learning. To overcome the specific difficulties of detecting minute objects in high-resolution aerial imagery, we curate a significant UAV dataset for training and rigorously evaluate the model's generalization capabilities. Specifically, we validate performance across three distinct test sets acquired at different temporal intervals, thereby assessing robustness against varying imaging conditions. Our empirical results demonstrate that transfer learning not only facilitates the rapid convergence of object detection models in agricultural contexts but also yields consistent performance despite domain shifts in image acquisition.