Author name cluster

Longguang Wang

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

3 papers

2 author rows

AAAI Conference 2025 Conference Paper

AIQViT: Architecture-Informed Post-Training Quantization for Vision Transformers

Runqing Jiang
Ye Zhang
Longguang Wang
Pengpeng Yu
Yulan Guo

Post-training quantization (PTQ) has emerged as a promising solution for reducing the storage and computational cost of vision transformers (ViTs). Recent advances primarily target at crafting quantizers to deal with peculiar activations characterized by ViTs. However, most existing methods underestimate the information loss incurred by weight quantization, resulting in significant performance deterioration, particularly in low-bit cases. Furthermore, a common practice in quantizing post-Softmax activations of ViTs is to employ logarithmic transformations, which unfortunately prioritize less informative values around zero. This approach introduces additional redundancies, ultimately leading to suboptimal quantization efficacy. To handle these, this paper proposes an innovative PTQ method tailored for ViTs, termed AIQViT (Architecture-Informed Post-training Quantization for ViTs). First, we design an architecture-informed low-rank compensation mechanism, wherein learnable low-rank weights are introduced to compensate for the degradation caused by weight quantization. Second, we design a dynamic focusing quantizer to accommodate the unbalanced distribution of post-Softmax activations, which dynamically selects the most valuable interval for higher quantization resolution. Extensive experiments on five vision tasks, including image classification, object detection, instance segmentation, point cloud classification, and point cloud part segmentation, demonstrate the superiority of AIQViT over state-of-the-art PTQ methods.

PDF Details DOI

JBHI Journal 2025 Journal Article

Cross-Scale Texture Supplementation for Reference-based Medical Image Super-Resolution

Yinghua Li
Weiao Hao
Hao Zeng
Longguang Wang
Jian Xu
Sidheswar Routray
Rutvij H. Jhaveri
Thippa Reddy Gadekallu

Magnetic Resonance Imaging (MRI) is a widely used medical imaging technique, but its resolution is often limited by acquisition time constraints, potentially compromising diagnostic accuracy. Reference-based Image Super-Resolution (RefSR) has shown promising performance in addressing such challenges by leveraging external high-resolution (HR) reference images to enhance the quality of low-resolution (LR) images. The core objective of RefSR is to accurately establish correspondences between the reference HR image and the LR images. In pursuit of this objective, this paper develops a Self-rectified Texture Supplementation network for RefSR (STS-SR) to enhance fine details in MRI images and support the expanding role of autonomous AI in healthcare. Our network comprises a texture-specified selfrectified feature transfer module and a cross-scale texture complementary network. The feature transfer module employs highfrequency filtering to facilitate the network concentrating on fine details. To better exploit the information from both the reference and LR images, our cross-scale texture complementary module incorporates the All-ViT and Swin Transformer layers to achieve feature aggregation at multiple scales, which enables high-quality image enhancement that is critical for autonomous AI systems in healthcare to make accurate decisions. Extensive experiments are performed across various benchmark datasets. The results validate the effectiveness of our method and demonstrate that the method produces state-of-the-art performance as compared to existing approaches. This advancement enables autonomous AI systems to utilize high-quality MRI images for more accurate diagnostics and reliable predictions.

Details DOI

ICLR Conference 2024 Conference Paper

ACRF: Compressing Explicit Neural Radiance Fields via Attribute Compression

Guangchi Fang
Qingyong Hu
Longguang Wang
Yulan Guo

In this work, we study the problem of explicit NeRF compression. Through analyzing recent explicit NeRF models, we reformulate the task of explicit NeRF compression as 3D data compression. We further introduce our NeRF compression framework, Attributed Compression of Radiance Field (ACRF), which focuses on the compression of the explicit neural 3D representation. The neural 3D structure is pruned and converted to points with features, which are further encoded using importance-guided feature encoding. Furthermore, we employ an importance-prioritized entropy model to estimate the probability distribution of transform coefficients, which are then entropy coded with an arithmetic coder using the predicted distribution. Within this framework, we present two models, ACRF and ACRF-F, to strike a balance between compression performance and encoding time budget. Our experiments, which include both synthetic and real-world datasets such as Synthetic-NeRF and Tanks&Temples, demonstrate the superior performance of our proposed algorithm.

Details