AI 2023: Advances in Artificial Intelligence : 36th Australasian Joint Conference on Artificial Intelligence, AI 2023, Brisbane, QLD, Australia, November 28–December 1, 2023, Proceedings, Part I / edited by Tongliang Liu, Geoff Webb, Lin Yue, Dadong Wang.

Format:

Book

Contributor:

Liu, Tongliang, editor.

Series:

Lecture Notes in Artificial Intelligence, 2945-9141 ; 14471

Language:

English

Subjects (All):

Artificial intelligence.

Computer networks.

Data mining.

Application software.

Computer vision.

Artificial Intelligence.

Computer Communication Networks.

Data Mining and Knowledge Discovery.

Computer and Information Systems Applications.

Computer Vision.

Local Subjects:

Artificial Intelligence.

Computer Communication Networks.

Data Mining and Knowledge Discovery.

Computer and Information Systems Applications.

Computer Vision.

Physical Description:

1 online resource (574 pages)

Edition:

1st ed. 2024.

Place of Publication:

Singapore : Springer Nature Singapore : Imprint: Springer, 2024.

Summary:

This two-volume set LNAI 14471-14472 constitutes the refereed proceedings of the 36th Australasian Joint Conference on Artificial Intelligence, AI 2023, held in Brisbane, QLD, Australia during November 28 – December 1, 2023. The 23 full papers presented together with 59 short papers were carefully reviewed and selected from 213 submissions. They are organized in the following topics: computer vision; deep learning; machine learning and data mining; optimization; medical AI; knowledge representation and NLP; explainable AI; reinforcement learning; and genetic algorithm. .

Contents:

Intro

Preface

Organization

Contents - Part I

Contents - Part II

Computer Vision

Multi-graph Laplacian Feature Mapping Incorporating Tag Information for Image Annotation

1 Introduction

2 Related Work

3 Propoesd Method

3.1 Multi-graph Laplacian Incorporating Tag Information

3.2 Tag Graph Laplacian with Visual Content

3.3 Loss Function and Objective Function

4 Optimization

5 Experimental Results

5.1 Experiment Settings

5.2 Experimental Performance

5.3 The Analysis Parameters

6 Conclusion

References

Short-Term Solar Irradiance Forecasting from Future Sky Images Generation

2 Related Works

3 Method

3.1 Nowcasting Model

3.2 Image Prediction Model

3.3 The Forecasting Framework

4 Experiments and Results

4.1 Datasets

4.2 Evaluate Metrics, Data Processing and Hyper-parameters

4.3 Nowcasting Results

4.4 Forecasting Results

5 Conclusion

No Token Left Behind: Efficient Vision Transformer via Dynamic Token Idling

3 Methods

3.1 Preliminaries

3.2 Token Selection and Idling

3.3 Token Cut Loss

3.4 Finetuning

4 Experiments

4.1 Implementation Settings

4.2 Results

4.3 Analysis of Token Cut Loss

4.4 Analysis of Token Idle Strategy

Story Sifting Using Object Detection Techniques

2 Background and Related Work

3 Approach

3.1 Recasting Story Sifting as Object Detection

3.2 Representing Story Arcs as Images

3.3 Choice of YOLOv5 Model

4 Model Development

5 Evaluating Model Performance

5.1 Model Performance

6 Evaluating Time Efficiency

7 Detection from a Virtual Storyworld Environment

8 Discussion

9 Conclusion

References.

SimMining-3D: Altitude-Aware 3D Object Detection in Complex Mining Environments: A Novel Dataset and ROS-Based Automatic Annotation Pipeline

2 Related Study

3 New Dataset: SimMining3D

3.1 Data Collection at Simulated Environment

3.2 Automatic Annotation

4 Perception: Baseline Experiment

4.1 Experimental Setup

4.2 Results and Discussion

Oyster Mushroom Growth Stage Identification: An Exploration of Computer Vision Technologies

3 The Monitoring System

3.1 The Problem

3.2 The Label Map

4 Empirical Studies

4.1 Settings

4.2 Performances

5 Conclusion and Future Works

Handling Heavy Occlusion in Dense Crowd Tracking by Focusing on the Heads

3 Methodology

3.1 Framework Overview

3.2 Anchor-Free Head-Body Detection

3.3 Joint SimOTA

3.4 Tracking Framework

3.5 Loss Function

3.6 Training Details

4.1 MOT Challenge

4.2 Qualitative Result on MOT20

4.3 Ablation Study on Joint SimOTA

4.4 Crowdhuman

SAR2EO: A High-Resolution Image Translation Framework with Denoising Enhancement

2.1 GAN

2.2 Image-to-Image Translation

3 Proposed Method

3.1 Preliminary: Pix2pixHD

3.2 SAR and EO Images

3.3 Denoising Enhanced SAR2EO Framework

4.1 Dataset

4.2 Metrics

4.3 Implementation Details

4.4 Main Results

4.5 Ablation Studies

A New Perspective of Weakly Supervised 3D Instance Segmentation via Bounding Boxes

2.1 Fully Supervised Method

2.2 Weakly Supervised Method

3.1 Problem Description

3.2 Cluster-Based Candidate Points Filtering.

3.3 Smallest-Box Heuristic

4 Experiment

4.1 Implementation Details

4.2 Dataset

4.3 Evaluation Metrics and Experiment Results

4.4 Ablation Study

4.5 Robustness

Large-Kernel Attention Network with Distance Regression and Topological Self-correction for Airway Segmentation

2 Method

2.1 Network Architecture

2.2 Prediction Head

2.3 Implementation Details

3 Experimental Results

3.1 Metrics

3.2 Comparison with Other Methods

3.3 Ablation Study

4 Conclusion

Deep Learning

WeightRelay: Efficient Heterogeneous Federated Learning on Time Series

2.1 Deep Learning for Time Series Classification

2.2 Federated Learning on Heterogeneous Devices

3 Motivation

4 Weight Relay

4.1 Heterogeneous Models

4.2 Weight Alignment

5 Analysis of Weight Relay

5.1 Consistency Proof for the Alignment

5.2 Macro Explanation of the Training Acceleration

5.3 Micro Explanation of the Training Acceleration

6 Experiment

6.1 Benchmarks

6.2 Evaluation Criteria

6.3 Experiment Setup

6.4 Experiment Result

7 Conclusion

Superpixel Attack

2 Preliminaries

2.1 Problem Definition

2.2 Related Work

3 Research on Update Areas

3.1 Update Areas of Existing Methods

3.2 Color Variance of Update Areas

3.3 Compactness of Update Areas

3.4 Superpixel Calculated by SLIC

3.5 Analysis of Color Variance and Compactness

4 Superpixel Attack

4.1 Update Areas Using Superpixels

4.2 Procedure of Versatile Search

5 Experiments

Cross Domain Pulmonary Nodule Detection Without Source Data

2.1 Feature Extractor Adaptation

2.2 Detection Head Adaptation

3 Experiments.

3.1 Benchmark and Evaluation

3.2 Implementation Details

3.3 Results

4 Related Works

3RE-Net: Joint Loss-REcovery and Super-REsolution Neural Network for REal-Time Video

3 Model Design

Neural Networks in Forecasting Financial Volatility

3 Experimental Comparison of Forecasting Models

3.1 Posing the Problem as a Shared Task

3.2 Methods

3.3 Result Evaluation and Analysis

4 Discussion

CLIP-Based Composed Image Retrieval with Comprehensive Fusion and Data Augmentation

2.1 Composed Image Retrieval

2.2 Vision-Language Pre-training

3.1 Problem Formulation

3.2 CLIP-CD

4.1 Datasets and Metrics

4.2 Implementation Details

4.3 Performance Comparison

4.5 Case Study

5 Conclusions

LiDAR Inpainting of UAV Based 3D Point Cloud Using Supervised Learning

3 Preliminaries

4 Problem Definition

5 Methodology

5.1 Simulator

5.2 Extracting Individual Point Clouds

5.3 Point Cloud Inpainting Model

5.4 Inpainting Complete Environments

6 Experimental Results

7 Conclusion and Future Work

A Sampling Method for Performance Predictor Based on Contrastive Learning

2 Background

2.1 Contrastive Learning

2.2 Graph Data Sampling Methods

3.1 Architecture Augmentation

3.2 Architecture Maximal Agreement

4.1 Overall Performance

4.2 Performance Evaluation in NAS Datasets

4.3 Ablation Study

AdaptMatch: Adaptive Consistency Regularization for Semi-supervised Learning with Top-k Pseudo-labeling and Contrastive Learning

2.1 Consistency Regularization

2.2 Contrastive Learning

3 Our Approach: AdaptMatch

3.1 Data Augmentation

3.2 Top-k Label Guessing

3.3 Contrastive Learning

3.4 Summarization of the Framework

4.1 Datasets and Experimental Setup

4.2 Main Results

5 Ablation Study

Estimation of Unmasked Face Images Based on Voice and 3DMM

2 Related Research

2.1 Studies on Mask Removal

2.2 Studies on Estimating Facial Shape from Voice

2.3 3D Morphable Model (3DMM)

3.1 Overview of the Proposal Method

3.2 Extraction of Voice Embedding

3.3 Combining Voice Embedding and Intermediate Features

3.4 Training of Multitasking Module

4.2 Training Details

4.3 Qualitative Evaluation

4.4 Quantitative Evaluation

5 Discussion

5.1 On Qualitative Evaluation

5.2 On Quantitative Evaluation

6 Conclusion and Future Works

Aging Contrast: A Contrastive Learning Framework for Fish Re-identification Across Seasons and Years

2.1 Deep Learning for Fish Recognition

3 Dataset

4 Proposed Method

4.1 Segmentation and Feature Extraction

4.2 Aging Contrast Framework

Spatial Bottleneck Transformer for Cellular Traffic Prediction in the Urban City

3 Problem Formulation

4 Methodology

4.1 Spatial Bottleneck Transformer

4.2 ST-InducedTran Model

5.1 Dataset

5.2 Baseline

5.3 Implementation Details

5.4 Evaluation Metrics.

6 Results and Discussion.

Other Format:

Print version: Liu, Tongliang AI 2023: Advances in Artificial Intelligence

ISBN:

981-9983-88-6