Search

Graph computing has emerged as a focal point in recent research across various fields, including the realm of 3D instance segmentation, where it aids in detecting and segmenting objects within volumetric data. Our study introduces GraphFusion, a state-of-the-art network that harnesses the power of graph computing to enhance the segmentation of 3D point clouds. GraphFusion is equipped with a Multi-Level Semantic Aggregation Module, architectured akin to a graph, to capture comprehensive features from 3D point clouds. Utilizing graph-based methodologies, this module proficiently aggregates multi-scale semantic information, illuminating insights from both global and local contexts. Additionally, our Parallel Feature Fusion Transformer Module leverages graph-transformer techniques to intricately process complex spatial relationships within point clouds, culminating in a more cohesive feature representation. Rigorous experiments on the ScanNetv2 dataset affirm the dominance of GraphFusion, which eclipses current methods by 2.2% in mean Average Precision (mAP) on the hidden test set. The model's code is accessible at https://github.com/3171228612/GraphFusion. [ABSTRACT FROM AUTHOR]

Aspect Sentiment Triplet Extraction is an emerging and challenging task that attempts to present a complete picture of aspect-based sentiment analysis. Prior research efforts mostly leverage various tagging schemes to extract the three elements in a triplet. However, these methods fail to explicitly model the complicated relations between aspects and opinions and the boundaries of multi-word aspects and opinions. In this paper, we propose a bi-syntax guided transformer network in an end-to-end manner to address these challenges. Firstly, we devise three types of representations, including sequence distance representation, constituency distance representation, and dependency distance representation, to learn the comprehensive language representation. Specifically, sequence distance representation utilizes sequence distance between words to enhance the contextual representation. Constituency distance representation adopts constituency distance between words in a constituency tree to capture the intra-span relation between words. Dependency distance representation employs dependency distance between words in a dependency tree to capture the long-distance relation between aspects and opinions. Extensive experiments are conducted on four benchmark datasets to validate the effectiveness of our method. The results demonstrate that the proposed approach achieves better performance than baseline methods. We conduct further detailed analysis to demonstrate that our method effectively handles multi-word terms and overlapping triplets. [ABSTRACT FROM AUTHOR]

Blind Face Restoration (BFR) aims to generate high-quality face images from low-quality inputs. However, existing BFR methods often use private datasets for training and evaluation, making it challenging for future approaches to compare fairly. To address this issue, we introduce two benchmark datasets, BFRBD128 and BFRBD512, for evaluating state-of-the-art methods in five scenarios: blur, noise, low resolution, JPEG compression artifacts, and full degradation. We use seven standard quantitative metrics and two task-specific metrics, AFLD and AFICS. Additionally, we propose an efficient baseline model called Swin Transformer U-Net (STUNet), which outperforms state-of-the-art methods in various BFR tasks. The codes, datasets, and trained models are publicly available at: https://github.com/bitzpy/Blind-Face-Restoration-Benchmark-Datasets-and-a-Baseline-Model. [ABSTRACT FROM AUTHOR]

The task of human parsing aims to segment the human body into different semantic regions. Despite advancements in this field, there are still two issues with current works: boundary indistinction and parsing inconsistency. In this paper, we investigate how to utilize structural information and auxiliary information to jointly solve the above two problems. Drawing inspiration from Transformer architecture, a Boundary-guided Part Reasoning Network (BPRNet) is proposed to combine edge information and associated semantics of body parts for human parsing. Specifically, we design a part representation module to represent human body parts as part features. Based on the Transformer decoder, a multi-head self-attention is used to capture the semantic correlation between the human body. Moreover, we propose a boundary-guided module consisting of absolute boundary attention and reinforced boundary attention. They take advantage of edge information and multi-scale image features to jointly constrain cross-attention to extract global features. Experiments and corresponding results on three public datasets show that the proposed method performs favorably against the state-of-the-art methods. [ABSTRACT FROM AUTHOR]