IEEE

Efficient Encoding and Decoding of Voxelized Models for Machine Learning-Based Applications

Abstract: Point clouds have become a popular training data for many practical applications of machine learning in the fields of environmental modeling and precision agriculture. In order to reduce ...
GitHub

Add GPU-Accelerated Video Decoding for VideoMAE, TimeSformer, and Similar Models

I’d like to propose adding a lightweight, GPU-friendly video decoding utility to Transformers. Right now, models like VideoMAE and TimeSformer depend on PIL or OpenCV for frame extraction, which keeps ...
IEEE

DUSA-UNet: Dual Sparse Attentive U-Net for Multiscale Road Network Extraction

Abstract: The challenges of road network segmentation demand an algorithm capable of adapting to the sparse and irregular shapes, as well as the diverse context, which often leads traditional encoding ...
Frontiers

LEM-UNet: an edge-guided network for 3D multimodal images segmentation in focal cortical dysplasia

1 School of Information and Communication Engineering, Beijing University of Posts and Telecommunications, Beijing, China 2 Pediatric Epilepsy Center, Peking University First Hospital, Beijing, China ...
GitHub

Free Draft-and-Verification: Toward Lossless Parallel Decoding for Diffusion Large Language Models

We propose FreeDave (Free Draft-and-Verification), a fast sampling algorithm for diffusion language models, which achieves lossless parallel decoding via a pipeline of parallel-decoded candidate ...