DAKD: Data Augmentation and Knowledge Distillation using Diffusion Models for SAR Oil Spill Segmentation

2024-12-11

arXiv.org

AI Summary

To address the dual challenges of scarce annotated data and speckle noise interference in synthetic aperture radar (SAR) oil-spill segmentation, this work proposes a diffusion-augmented knowledge distillation framework (DAKD) and a dedicated segmentation network, SAROSS-Net. We introduce SAR-JointNet—the first joint SAR image–label generative diffusion model—enabling high-fidelity, co-synthesized image-label pairs from limited samples. Further, we design a cross-modal soft-label distillation mechanism and a context-aware feature transfer module to enhance robustness against speckle noise and improve feature discriminability. Evaluated on public SAR oil-spill datasets, our approach achieves a 6.2% improvement in mean Intersection-over-Union (mIoU) over state-of-the-art methods, demonstrates superior generalization, and yields precisely spatially aligned generated images and labels—effectively mitigating the constraints imposed by both data scarcity and noise corruption.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Search and Optimization: Distributed Search

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Web data integration and cleaning

Abstract

Oil spills in the ocean pose severe environmental risks, making early detection essential. Synthetic aperture radar (SAR) based oil spill segmentation offers robust monitoring under various conditions but faces challenges due to the limited labeled data and inherent speckle noise in SAR imagery. To address these issues, we propose (i) a diffusion-based Data Augmentation and Knowledge Distillation (DAKD) pipeline and (ii) a novel SAR oil spill segmentation network, called SAROSS-Net. In our DAKD pipeline, we present a diffusion-based SAR-JointNet that learns to generate realistic SAR images and their labels for segmentation, by effectively modeling joint distribution with balancing two modalities. The DAKD pipeline augments the training dataset and distills knowledge from SAR-JointNet by utilizing generated soft labels (pixel-wise probability maps) to supervise our SAROSS-Net. The SAROSS-Net is designed to selectively transfer high-frequency features from noisy SAR images, by employing novel Context-Aware Feature Transfer blocks along skip connections. We demonstrate our SAR-JointNet can generate realistic SAR images and well-aligned segmentation labels, providing the augmented data to train SAROSS-Net with enhanced generalizability. Our SAROSS-Net trained with the DAKD pipeline significantly outperforms existing SAR oil spill segmentation methods with large margins.

Problem

• Addresses scarcity of labeled data in SAR oil spill segmentation.
• Proposes diffusion-based data augmentation with knowledge transfer.
• Enhances segmentation models using generated images and soft labels.

Innovation

• Diffusion-based data augmentation generates SAR images.
• Soft labels provide rich class probability distributions.
• SNR-based balancing aligns noise in diffusion models.