QGait: Toward Accurate Quantization for Gait Recognition

A new study has introduced a neural architecture search (NAS) framework aimed at automating the design of quantum autoencoders using a genetic algorithm. This development addresses the complexities involved in selecting gates, arranging circuit layers, and tuning parameters for effective quantum circuit architectures, which are essential for compressing high-dimensional quantum and classical data.

A recent study has proposed a novel approach for the unsupervised and source-free ranking of biomedical segmentation models, addressing the significant challenge of model selection in the biomedical community where data annotation is costly and time-consuming. This method builds on previous research linking model generalization and consistency under perturbation, aiming to facilitate the adoption of pre-trained models without the need for target dataset labels.

A new study introduces a periodicity-enforced neural network approach for designing Deterministic Lateral Displacement (DLD) devices, which are crucial for liquid biopsy in cancer detection by effectively separating circulating tumor cells from blood samples. This method enhances the design process by addressing the limitations of traditional computational simulations, particularly in managing periodic boundary conditions.

The study introduces GeeSanBhava, a comprehensive dataset of Sinhala song comments sourced from YouTube, which has been meticulously tagged using Russell's Valence-Arousal model by three independent annotators, achieving a high inter-annotator agreement of 84.96%. This dataset highlights the emotional profiles associated with different songs, emphasizing the significance of comment-based emotion mapping.

Recent advancements in optical coherence tomography angiography (OCTA) have led to a novel approach for robust detection of retinal neovascularization (RNV) in widefield imaging. This development is crucial for timely intervention in diabetic retinopathy, a condition that can lead to vision loss. The new model reframes RNV identification, enhancing detection capabilities beyond conventional methods.

SemanticStitch is a newly introduced deep learning framework designed to enhance image coherence by incorporating semantic information of foreground objects during the seam carving process. This innovation addresses common challenges in image stitching, such as misalignments and visual discrepancies caused by varying capture angles and object movements.

A new framework named CroTad has been introduced for online trajectory anomaly detection, addressing critical challenges in Intelligent Transportation Systems (ITS). This method utilizes contrastive reinforcement learning to detect anomalies in sub-trajectories without relying on predefined thresholds, enhancing adaptability in real-world applications.

A new framework for stable coreset selection in deep learning has been proposed, addressing challenges in training efficiency and representativeness due to loss landscape mismatches. This framework connects posterior sampling with loss landscapes, enhancing coreset selection even in high data corruption scenarios.