Hear What Matters! Text-conditioned Selective Video-to-Audio Generation

A new end-to-end framework for multi-person 2D pose estimation in videos has been introduced, eliminating the reliance on heuristic operations that limit accuracy and efficiency. This framework, named Pose-Aware Video transformEr Network (PAVE-Net), effectively associates individuals across frames, addressing the challenges of complex and overlapping trajectories in video data.

Recent advancements in dance generation have led to the development of a novel approach that utilizes a generative masked text-to-motion model to synthesize high-quality 3D dance motions. This method addresses significant challenges such as realism, dance-music synchronization, and motion diversity, while also enabling semantic motion editing capabilities.

A new framework for fast 3D surrogate modeling has been developed to enhance thermal management in data centers, focusing on real-time temperature predictions that are crucial for energy efficiency and sustainability. This approach utilizes a voxelized representation of the data center, integrating various operational parameters such as server workloads and HVAC settings.

A new paper introduces a context-enriched contrastive loss function aimed at improving the effectiveness of contrastive learning frameworks. This approach addresses the issue of information distortion that arises from augmented samples, which can lead to models over-relying on identical label information while neglecting positive pairs from the same image. The proposed method incorporates two convergence targets to enhance learning outcomes.

A new study has introduced a method for enhancing medical Vision-Language Models (VLMs) through momentum self-distillation, addressing the challenges posed by limited computing resources and the scarcity of detailed annotations in healthcare. This approach aims to improve the efficiency of training VLMs, allowing them to perform well even with small datasets or in zero-shot scenarios.

A new framework called Basis-Oriented Low-rank Transfer (BOLT) has been proposed to enhance the adaptation of large pre-trained models to unseen tasks with minimal additional training. This method focuses on extracting an orthogonal, task-informed spectral basis from existing fine-tuned models, allowing for efficient adaptation in both offline and online phases.

HouseLayout3D has been introduced as a benchmark for 3D layout estimation, addressing limitations of existing models that primarily rely on synthetic datasets. This new benchmark supports the estimation of layouts in complex multi-floor buildings, which are often overlooked in current methodologies.

The recent introduction of Trajectory Guided Dataset Distillation (TGDD) aims to enhance dataset distillation by reformulating distribution matching as a dynamic alignment process throughout the model's training trajectory. This method captures evolving semantics by aligning feature distributions between synthetic and original datasets, while also implementing a distribution constraint to minimize class overlap.