SlowFast‑SCI introduces a dual‑speed deep‑unfolding framework for spectral compressive imaging that combines a slow, pre‑trained backbone with a fast, test‑time adaptation stage. The slow phase distills a priors‑based model into a compact fast‑unfolding network, while the fast phase embeds lightweight modules that self‑supervise at test time without retraining the backbone. This design yields significant reductions in parameters and FLOPs, improves out‑of‑distribution PSNR by up to 5.79 dB, and accelerates adaptation four‑fold, all while remaining modular enough to integrate with any existing deep‑unfolding system.
arXiv:2507. 06764v5 Announce Type: replace-cross Abstract: In this work, we propose Fast Equivariant Imaging (FEI), a novel unsupervised learning framework to rapidly and efficiently train deep imaging networks without ground-truth data.
HyperVision introduces the first ground‑based hyperspectral pre‑trained backbone, addressing challenges of varying spectral configurations, limited annotations, and dataset diversity. It employs a channel‑adaptive dynamic embedding to unify heterogeneous inputs, a multi‑source pseudo‑labeling strategy combining SAM2 spatial cues with HyperFree spectral details, and cross‑modal knowledge distillation from a pre‑trained RGB vision model. Trained on 15k images from 26 datasets, HyperVision achieves significant improvements—up to 16.3% relative gain in hyperspectral semantic segmentation, 2.1% in object tracking AUC, and 35.5% reduction in salient object detection MAE—while requiring only head‑only adaptation.
arXiv:2603. 23647v2 Announce Type: replace-cross Abstract: In fluorescence microscopy, spectral unmixing aims to recover individual fluorophore concentrations from spectral images that capture mixed fluorophore emissions.
arXiv:2609.14391v1 Announce Type: new Abstract: Compressed sensing (CS) reconstructs images from highly limited measurements, but existing deep unfolding methods are typically driven by first-order o...
WildRelight is the first in-the-wild dataset designed to evaluate single-image relighting models, featuring high-resolution outdoor scenes captured under strictly aligned, temporally varying natural illuminations paired with high-dynamic-range environment maps. The benchmark demonstrates that state-of-the-art models trained on synthetic data suffer severe domain shifts when applied to real-world imagery. Leveraging the dataset’s temporal structure, the authors introduce a physics-guided inference framework combining Diffusion Posterior Sampling with Temporal Sampling-Aware Test-Time Adaptation, enabling synthetic models to self-supervise and align with real-world statistics on-the-fly.
arXiv:2603. 18846v3 Announce Type: replace-cross Abstract: Foundation models are used to extract transferable representations from large amounts of unlabeled data, typically via self-supervised learning (SSL).
arXiv:2607.01630v2 Announce Type: replace Abstract: Dynamic expansion methods for class-incremental learning (CIL) protect task-specific knowledge by growing dedicated tokens or subnetworks, yet our...
arXiv:2607. 04603v1 Announce Type: cross Abstract: Infrared small target detection (IRSTD) aims to identify long distance small targets from complex infrared backgrounds, and is a fundamental task in remote sensing.
The paper introduces a self‑supervised approach for Poisson inverse imaging problems that leverages frozen CLIP RN50 features as a lightweight prior within an ADMM‑inspired unrolled solver. By decoupling data‑consistency from the prior and using a parameter‑efficient decoder, the method adapts foundation vision representations without extensive fine‑tuning. Experiments on Poisson CFA demosaicing and deblurring demonstrate competitive image quality, enhanced robustness to dataset and acquisition shifts, and self‑supervised performance close to that of supervised training.
arXiv:2403. 18026v3 Announce Type: replace-cross Abstract: High-throughput imaging is often constrained by a trade-off between acquisition speed and image quality.
SIMPLER is a pre‑fine‑tuning method that reduces inference and deployment costs for Earth Observation foundation models by pruning redundant layers. It uses layer‑wise representation similarity on unlabeled task data to identify and remove up to 79% of parameters without requiring gradients, magnitude heuristics, or hyperparameter tuning. Experiments on Prithvi‑EO‑2, TerraMind, and ImageNet‑pretrained ViT‑MAE show that SIMPLER retains 94% of baseline performance while achieving 2.1× faster training and 2.6× faster inference.
arXiv:2608. 15972v1 Announce Type: cross Abstract: Synchronized camera and wireless measurements observe the same scene through different physical channels.