SPWOOD: Sparse Partial Weakly-Supervised Oriented Object Detection

TL;DR

SPWOOD tackles the high annotation cost in oriented object detection for remote sensing by introducing a Sparse Partial Weakly-Supervised framework that combines a Sparse Annotation learning-enabled SOS-Student, a Multi-level Pseudo-label Filtering strategy, and a unified sparse annotation approach. The two-branch training pipeline leverages abundant unlabeled data and limited sparse-weak annotations to learn object orientation and scale while generating reliable pseudo-labels through EMA teachers. Across DOTA v1.0/v1.5 and DIOR, SPWOOD delivers significant improvements over existing semi-supervised and weakly supervised methods, validating its effectiveness and cost-efficiency for large-scale remote sensing applications. The work further contributes methodological innovations such as a tailored SAL loss, symmetry-based orientation learning, Gaussian overlap and Voronoi watershed scale losses, and a data-distribution-preserving sparse sampling strategy, paving the way for practical deployment with diverse annotation formats.

Abstract

A consistent trend throughout the research of oriented object detection has been the pursuit of maintaining comparable performance with fewer and weaker annotations. This is particularly crucial in the remote sensing domain, where the dense object distribution and a wide variety of categories contribute to prohibitively high costs. Based on the supervision level, existing oriented object detection algorithms can be broadly grouped into fully supervised, semi-supervised, and weakly supervised methods. Within the scope of this work, we further categorize them to include sparsely supervised and partially weakly-supervised methods. To address the challenges of large-scale labeling, we introduce the first Sparse Partial Weakly-Supervised Oriented Object Detection framework, designed to efficiently leverage only a few sparse weakly-labeled data and plenty of unlabeled data. Our framework incorporates three key innovations: (1) We design a Sparse-annotation-Orientation-and-Scale-aware Student (SOS-Student) model to separate unlabeled objects from the background in a sparsely-labeled setting, and learn orientation and scale information from orientation-agnostic or scale-agnostic weak annotations. (2) We construct a novel Multi-level Pseudo-label Filtering strategy that leverages the distribution of model predictions, which is informed by the model's multi-layer predictions. (3) We propose a unique sparse partitioning approach, ensuring equal treatment for each category. Extensive experiments on the DOTA and DIOR datasets show that our framework achieves a significant performance gain over traditional oriented object detection methods mentioned above, offering a highly cost-effective solution. Our code is publicly available at https://github.com/VisionXLab/SPWOOD.

Figures (3)

• Figure 1: Current oriented object detection methods are predominantly classified into five categories. Compared to the aforementioned approaches, our proposed Sparse Partial Weakly-supervised Oriented Object Detection (SPWOOD) distinguished with minimal annotation requirements.
• Figure 2: The illustration of the Sparse Partial Weakly-supervised Oriented Object Detection (SPWOOD). The Sparse-annotation-Orientation-and-Scale-aware Student (SOS-Student) identifies hard negatives and learn the scale and angle information from sparse weak annotation data. The Multi-level Pseudo-labels Filtering (MPF) mechanism acquires ability from student through EMA algorithm and selects high-quality pseudo-labels for student module's training.
• Figure 3: Qualitative results showing the qualities of the detection performance.