The early detection of glaucoma is essential in preventing visual impairment. Artificial intelligence (AI) can be used to analyze color fundus photographs (CFPs) in a cost-effective manner, making glaucoma screening more accessible. While AI models for glaucoma screening from CFPs have shown promising results in laboratory settings, their performance decreases significantly in real-world scenarios due to the presence of out-of-distribution and low-quality images. To address this issue, we propose the Artificial Intelligence for Robust Glaucoma Screening (AIROGS) challenge. This challenge includes a large dataset of around 113,000 images from about 60,000 patients and 500 different screening centers, and encourages the development of algorithms that are robust to ungradable and unexpected input data. We evaluated solutions from 14 teams in this paper and found that the best teams performed similarly to a set of 20 expert ophthalmologists and optometrists. The highest-scoring team achieved an area under the receiver operating characteristic curve of 0.99 (95% CI: 0.98-0.99) for detecting ungradable images on-the-fly. Additionally, many of the algorithms showed robust performance when tested on three other publicly available datasets. These results demonstrate the feasibility of robust AI-enabled glaucoma screening.

In this paper, we propose a Mask-Robust Inpainting Network (MRIN) approach to recover the masked areas of an image. Most existing methods learn a single model for image inpainting, under a basic assumption that all masks are from the same type. However, we discover that the masks are usually complex and exhibit various shapes and sizes at different locations of an image, where a single model cannot fully capture the large domain gap across different masks. To address this, we learn to decompose a complex mask area into several basic types and recover the damaged image in a patch-wise manner with a type-specific generator. More specifically, our MRIN consists of a mask-robust agent and an adaptive patch generative network. The mask-robust agent contains a mask selector and a patch locator, which generates mask attention maps to select a patch at each step. Based on the predicted mask attention maps, the adaptive patch generative network inpaints the selected patch with the generators bank, so that it sequentially inpaints each patch with different patch generators according to its mask type. Extensive experiments demonstrate that our approach outperforms most state-of-the-art approaches on the Place2, CelebA, and Paris Street View datasets.

Accurately forecasting the future movements of surrounding vehicles is essential for safe and efficient operations of autonomous driving cars. This task is difficult because a vehicle's moving trajectory is greatly determined by its driver's intention, which is often hard to estimate. By leveraging attention mechanisms along with long short-term memory (LSTM) networks, this work learns the relation between a driver's intention and the vehicle's changing positions relative to road infrastructures, and uses it to guide the prediction. Different from other state-of-the-art solutions, our work treats the on-road lanes as non-Euclidean structures, unfolds the vehicle's moving history to form a spatio-temporal graph, and uses methods from Graph Neural Networks to solve the problem. Not only is our approach a pioneering attempt in using non-Euclidean methods to process static environmental features around a predicted object, our model also outperforms other state-of-the-art models in several metrics. The practicability and interpretability analysis of the model shows great potential for large-scale deployment in various autonomous driving systems in addition to our own.

Path planning is a key component in motion planning for autonomous vehicles. A path specifies the geometrical shape that the vehicle will travel, thus, it is critical to safe and comfortable vehicle motions. For urban driving scenarios, autonomous vehicles need the ability to navigate in cluttered environment, e.g., roads partially blocked by a number of vehicles/obstacles on the sides. How to generate a kinematically feasible and smooth path, that can avoid collision in complex environment, makes path planning a challenging problem. In this paper, we present a novel quadratic programming approach that generates optimal paths with resolution-complete collision avoidance capability.

Due to the loose authentication requirement between access points (APs) and clients, it is notoriously known that WLANs face long-standing threats such as rogue APs and network freeloading. Take the rogue AP problem as an example, unfortunately encryption alone does not provide authentication. APs need to be equipped with certificates that are trusted by clients ahead of time. This requires either the presence of PKI for APs or other forms of pre-established trust (e.g., distributing the certificates offline), none of which is widely used. Before any strong security solution is deployed, we still need a practical solution that can mitigate the problem. In this paper, we explore a non-cryptographic solution that is readily deployable today on end hosts (e.g., smartphones and laptops) without requiring any changes to the APs or the network infrastructure. The solution infers the Carrier Frequency Offsets (CFOs) of wireless devices from Channel State Information (CSI) as their hardware fingerprints without any special hardware requirement. CFO is attributed to the oscillator drift, which is a fundamental physical property that cannot be manipulated easily and remains fairly consistent over time but varies significantly across devices. The real experiments on 23 smartphones and 34 APs (with both identical and different brands) in different scenarios demonstrate that the detection rate could exceed 94%.

In today’s business world, firms are facing pressures to reduce costs, enhance productivity, and maintain quality in new product development. Past studies have provided evidence that modularity can enhance performance of new product development. However, real-life cases on how to implement the concept of modularity are limited. This article aims to propose a model for modularity implementation in the context of embedded software development. The model was applied in a software company in Hong Kong. Results from the case company provide evidence that the average reuse rate of software modules increased from 31% to 71% after the implementation, with productivity increasing by 258%, cost reducing by 70%, and quality increasing by 72%. The practical implications are finally discussed.

This study aims to examine the direct and mediating relationships between team diversity, individual and team learning, and individual and team innovation performance. Using a sample of 266 white-collar professionals in Hong Kong, we found that team diversity plays a substantial role in improving both individual and team learning. Individual learning and team learning are independent and positively related to individual and team innovation performance. Individual learning partially mediates the relationship between team diversity and team learning. Individual innovation performance is found to be a partial mediator of team learning and team innovation performance. This study contributes in understanding the unexplored mediating effects of individual learning on team learning, and the individual innovation performance on team innovation performance. Team diversity is found to be the main enabler in the overall model. It is hoped that the research paradigm of this article will generate further development of information systems and technology literature.

Software quality estimation is important yet difficult in software engineering studies. Historical quality datasets are used to build classification models for estimating fault-proneness. However, the missing values in the datasets severely affect the estimation ability and therefore, cause inconclusive decision-making. Among the single imputation approaches, k nearest neighbor (kNN) imputation is popular in empirical studies due to the relatively high accuracy. However, researchers are still calling for the optimal parameter setting of kNN imputation. In this study, a novel grey relational analysis based incomplete-instance kNN imputation is built for software quality data. An evaluation is conducted on four quality datasets with different simulated missingness scenarios to analyze the performance of the proposed imputation. The empirical results show that the proposed approach is superior to traditional kNN imputation and mean imputation in most cases. Moreover, the classification accuracy can be maintained or even improved by using this approach in classification tasks.

Software quality prediction is an important yet difficult problem in software project development and management. Historical datasets can be used to build models for software quality prediction. However, the missing data significantly affects the prediction ability of models in knowledge discovery. Instead of ignoring missing observations, we investigate and improve incomplete-case k-nearest neighbor based imputation. K-nearest neighbor imputation is widely applied but has rarely been improved to have the most appropriate parameter settings for each imputation. This work conducts imputation on four well-known software quality datasets to discover the impact of the new imputation method we proposed. We compare it with mean imputation and other commonly used versions of k-nearest neighbor imputation. The empirical results show that the proposed dynamic incomplete-case nearest neighbor imputation performs better when the missingness is completely at random or non-ignorable, regardless of the percentage of missing values.