Zhengwei Wang

Jing Xu, Wentao Shi, Pan Gao et al.

In recent works on semantic segmentation, there has been a significant focus on designing and integrating transformer-based encoders. However, less attention has been given to transformer-based decoders. We emphasize that the decoder stage is equally vital as the encoder in achieving superior segmentation performance. It disentangles and refines high-level cues, enabling precise object boundary delineation at the pixel level. In this paper, we introduce a novel transformer-based decoder called MUSTER, which seamlessly integrates with hierarchical encoders and consistently delivers high-quality segmentation results, regardless of the encoder architecture. Furthermore, we present a variant of MUSTER that reduces FLOPS while maintaining performance. MUSTER incorporates carefully designed multi-head skip attention (MSKA) units and introduces innovative upsampling operations. The MSKA units enable the fusion of multi-scale features from the encoder and decoder, facilitating comprehensive information integration. The upsampling operation leverages encoder features to enhance object localization and surpasses traditional upsampling methods, improving mIoU (mean Intersection over Union) by 0.4% to 3.2%. On the challenging ADE20K dataset, our best model achieves a single-scale mIoU of 50.23 and a multi-scale mIoU of 51.88, which is on-par with the current state-of-the-art model. Remarkably, we achieve this while significantly reducing the number of FLOPs by 61.3%. Our source code and models are publicly available at: https://github.com/shiwt03/MUSTER.

Yiheng Lin, Siyu Jiao, Xiaohan Lan et al.

Recent advances in Multimodal Large Language Models (MLLMs) and diffusion-based generative models have substantially improved prompt-driven image editing. However, scene text editing remains challenging, as it requires models to precisely modify textual content while preserving visual realism and non-target regions. Current open-source models still lag behind proprietary systems, largely due to the scarcity of high-quality training data and the lack of standardized benchmarks tailored to text editing. To address these challenges, we present TextSculptor, a comprehensive framework for data construction and evaluation of scene text editing. We first develop an automated data construction pipeline that combines text-aware image synthesis with programmatic text rendering and compositing. Based on this pipeline, we build TextSculpt-Data, a large-scale dataset containing 3.2M training samples, including 1.2M OCR-verified text-to-image samples and 2M paired text editing samples with naturally aligned source-target images and strong background consistency. We further introduce TextSculpt-Bench, a benchmark covering four fundamental text editing tasks: text addition, text replacement, text removal, and hybrid editing. To support reliable evaluation, we design a tailored protocol that measures text accuracy, visual quality, and background preservation through OCR-based text alignment, multimodal judgment, and background-region similarity. Extensive experiments show that TextSculptor improves open-source text editing performance and narrows the gap to proprietary models. The data and benchmark are available at https://github.com/linyiheng123/TextSculptor.

Zhengwei Wang, Qi She, Aljosa Smolic

Most of existing video action recognition models ingest raw RGB frames. However, the raw video stream requires enormous storage and contains significant temporal redundancy. Video compression (e.g., H.264, MPEG-4) reduces superfluous information by representing the raw video stream using the concept of Group of Pictures (GOP). Each GOP is composed of the first I-frame (aka RGB image) followed by a number of P-frames, represented by motion vectors and residuals, which can be regarded and used as pre-extracted features. In this work, we 1) introduce sampling the input for the network from partially decoded videos based on the GOP-level, and 2) propose a plug-and-play mulTi-modal lEArning Module (TEAM) for training the network using information from I-frames and P-frames in an end-to-end manner. We demonstrate the superior performance of TEAM-Net compared to the baseline using RGB only. TEAM-Net also achieves the state-of-the-art performance in the area of video action recognition with partial decoding. Code is provided at https://github.com/villawang/TEAM-Net.

Zhengwei Wang, Qi She, Aljosa Smolic

Spatial-temporal, channel-wise, and motion patterns are three complementary and crucial types of information for video action recognition. Conventional 2D CNNs are computationally cheap but cannot catch temporal relationships; 3D CNNs can achieve good performance but are computationally intensive. In this work, we tackle this dilemma by designing a generic and effective module that can be embedded into 2D CNNs. To this end, we propose a spAtio-temporal, Channel and moTion excitatION (ACTION) module consisting of three paths: Spatio-Temporal Excitation (STE) path, Channel Excitation (CE) path, and Motion Excitation (ME) path. The STE path employs one channel 3D convolution to characterize spatio-temporal representation. The CE path adaptively recalibrates channel-wise feature responses by explicitly modeling interdependencies between channels in terms of the temporal aspect. The ME path calculates feature-level temporal differences, which is then utilized to excite motion-sensitive channels. We equip 2D CNNs with the proposed ACTION module to form a simple yet effective ACTION-Net with very limited extra computational cost. ACTION-Net is demonstrated by consistently outperforming 2D CNN counterparts on three backbones (i.e., ResNet-50, MobileNet V2 and BNInception) employing three datasets (i.e., Something-Something V2, Jester, and EgoGesture). Codes are available at \url{https://github.com/V-Sense/ACTION-Net}.

Zhengwei Wang, Qi She, Tejo Chalasani et al.

Egocentric gestures are the most natural form of communication for humans to interact with wearable devices such as VR/AR helmets and glasses. A major issue in such scenarios for real-world applications is that may easily become necessary to add new gestures to the system e.g., a proper VR system should allow users to customize gestures incrementally. Traditional deep learning methods require storing all previous class samples in the system and training the model again from scratch by incorporating previous samples and new samples, which costs humongous memory and significantly increases computation over time. In this work, we demonstrate a lifelong 3D convolutional framework -- c(C)la(a)ss increment(t)al net(Net)work (CatNet), which considers temporal information in videos and enables lifelong learning for egocentric gesture video recognition by learning the feature representation of an exemplar set selected from previous class samples. Importantly, we propose a two-stream CatNet, which deploys RGB and depth modalities to train two separate networks. We evaluate CatNets on a publicly available dataset -- EgoGesture dataset, and show that CatNets can learn many classes incrementally over a long period of time. Results also demonstrate that the two-stream architecture achieves the best performance on both joint training and class incremental training compared to 3 other one-stream architectures. The codes and pre-trained models used in this work are provided at https://github.com/villawang/CatNet.

Zhengwei Wang, Qi She, Alan F. Smeaton et al.

Generative adversarial networks (GANs) are increasingly attracting attention in the computer vision, natural language processing, speech synthesis and similar domains. However, evaluating the performance of GANs is still an open and challenging problem. Existing evaluation metrics primarily measure the dissimilarity between real and generated images using automated statistical methods. They often require large sample sizes for evaluation and do not directly reflect human perception of image quality. In this work, we introduce an evaluation metric called Neuroscore, for evaluating the performance of GANs, that more directly reflects psychoperceptual image quality through the utilization of brain signals. Our results show that Neuroscore has superior performance to the current evaluation metrics in that: (1) It is more consistent with human judgment; (2) The evaluation process needs much smaller numbers of samples; and (3) It is able to rank the quality of images on a per GAN basis. A convolutional neural network (CNN) based neuro-AI interface is proposed to predict Neuroscore from GAN-generated images directly without the need for neural responses. Importantly, we show that including neural responses during the training phase of the network can significantly improve the prediction capability of the proposed model. Codes and data can be referred at this link: https://github.com/villawang/Neuro-AI-Interface.

Zhengwei Wang, Qi She, Tomas E. Ward

Generative adversarial networks (GANs) have been extensively studied in the past few years. Arguably their most significant impact has been in the area of computer vision where great advances have been made in challenges such as plausible image generation, image-to-image translation, facial attribute manipulation and similar domains. Despite the significant successes achieved to date, applying GANs to real-world problems still poses significant challenges, three of which we focus on here. These are: (1) the generation of high quality images, (2) diversity of image generation, and (3) stable training. Focusing on the degree to which popular GAN technologies have made progress against these challenges, we provide a detailed review of the state of the art in GAN-related research in the published scientific literature. We further structure this review through a convenient taxonomy we have adopted based on variations in GAN architectures and loss functions. While several reviews for GANs have been presented to date, none have considered the status of this field based on their progress towards addressing practical challenges relevant to computer vision. Accordingly, we review and critically discuss the most popular architecture-variant, and loss-variant GANs, for tackling these challenges. Our objective is to provide an overview as well as a critical analysis of the status of GAN research in terms of relevant progress towards important computer vision application requirements. As we do this we also discuss the most compelling applications in computer vision in which GANs have demonstrated considerable success along with some suggestions for future research directions. Code related to GAN-variants studied in this work is summarized on https://github.com/sheqi/GAN_Review.

Zhengwei Wang, Qi She, Alan F. Smeaton et al.

Generative adversarial networks (GANs) are increasingly attracting attention in the computer vision, natural language processing, speech synthesis and similar domains. Arguably the most striking results have been in the area of image synthesis. However, evaluating the performance of GANs is still an open and challenging problem. Existing evaluation metrics primarily measure the dissimilarity between real and generated images using automated statistical methods. They often require large sample sizes for evaluation and do not directly reflect human perception of image quality. In this work, we describe an evaluation metric we call Neuroscore, for evaluating the performance of GANs, that more directly reflects psychoperceptual image quality through the utilization of brain signals. Our results show that Neuroscore has superior performance to the current evaluation metrics in that: (1) It is more consistent with human judgment; (2) The evaluation process needs much smaller numbers of samples; and (3) It is able to rank the quality of images on a per GAN basis. A convolutional neural network (CNN) based neuro-AI interface is proposed to predict Neuroscore from GAN-generated images directly without the need for neural responses. Importantly, we show that including neural responses during the training phase of the network can significantly improve the prediction capability of the proposed model. Materials related to this work are provided at https://github.com/villawang/Neuro-AI-Interface.

Zhengwei Wang, Gang Wu

Graph Transformers (GTs) show considerable potential in graph representation learning. The architecture of GTs typically integrates Graph Neural Networks (GNNs) with global attention mechanisms either in parallel or as a precursor to attention mechanisms, yielding a local-and-global or local-to-global attention scheme. However, as the global attention mechanism primarily captures long-range dependencies between nodes, these integration schemes may suffer from information loss, where the local neighborhood information learned by GNN could be diluted by the attention mechanism. Therefore, we propose G2LFormer, featuring a novel global-to-local attention scheme where the shallow network layers use attention mechanisms to capture global information, while the deeper layers employ GNN modules to learn local structural information, thereby preventing nodes from ignoring their immediate neighbors. An effective cross-layer information fusion strategy is introduced to allow local layers to retain beneficial information from global layers and alleviate information loss, with acceptable trade-offs in scalability. To validate the feasibility of the global-to-local attention scheme, we compare G2LFormer with state-of-the-art linear GTs and GNNs on node-level and graph-level tasks. The results indicate that G2LFormer exhibits excellent performance while keeping linear complexity.

Eoin Brophy, Zhengwei Wang, Qi She et al.

Generative adversarial networks (GANs) studies have grown exponentially in the past few years. Their impact has been seen mainly in the computer vision field with realistic image and video manipulation, especially generation, making significant advancements. While these computer vision advances have garnered much attention, GAN applications have diversified across disciplines such as time series and sequence generation. As a relatively new niche for GANs, fieldwork is ongoing to develop high quality, diverse and private time series data. In this paper, we review GAN variants designed for time series related applications. We propose a taxonomy of discrete-variant GANs and continuous-variant GANs, in which GANs deal with discrete time series and continuous time series data. Here we showcase the latest and most popular literature in this field; their architectures, results, and applications. We also provide a list of the most popular evaluation metrics and their suitability across applications. Also presented is a discussion of privacy measures for these GANs and further protections and directions for dealing with sensitive data. We aim to frame clearly and concisely the latest and state-of-the-art research in this area and their applications to real-world technologies.

Qi She, Fan Feng, Qi Liu et al.

This report summarizes IROS 2019-Lifelong Robotic Vision Competition (Lifelong Object Recognition Challenge) with methods and results from the top $8$ finalists (out of over~$150$ teams). The competition dataset (L)ifel(O)ng (R)obotic V(IS)ion (OpenLORIS) - Object Recognition (OpenLORIS-object) is designed for driving lifelong/continual learning research and application in robotic vision domain, with everyday objects in home, office, campus, and mall scenarios. The dataset explicitly quantifies the variants of illumination, object occlusion, object size, camera-object distance/angles, and clutter information. Rules are designed to quantify the learning capability of the robotic vision system when faced with the objects appearing in the dynamic environments in the contest. Individual reports, dataset information, rules, and released source code can be found at the project homepage: "https://lifelong-robotic-vision.github.io/competition/".

Qi She, Fan Feng, Xinyue Hao et al.

The recent breakthroughs in computer vision have benefited from the availability of large representative datasets (e.g. ImageNet and COCO) for training. Yet, robotic vision poses unique challenges for applying visual algorithms developed from these standard computer vision datasets due to their implicit assumption over non-varying distributions for a fixed set of tasks. Fully retraining models each time a new task becomes available is infeasible due to computational, storage and sometimes privacy issues, while naïve incremental strategies have been shown to suffer from catastrophic forgetting. It is crucial for the robots to operate continuously under open-set and detrimental conditions with adaptive visual perceptual systems, where lifelong learning is a fundamental capability. However, very few datasets and benchmarks are available to evaluate and compare emerging techniques. To fill this gap, we provide a new lifelong robotic vision dataset ("OpenLORIS-Object") collected via RGB-D cameras. The dataset embeds the challenges faced by a robot in the real-life application and provides new benchmarks for validating lifelong object recognition algorithms. Moreover, we have provided a testbed of $9$ state-of-the-art lifelong learning algorithms. Each of them involves $48$ tasks with $4$ evaluation metrics over the OpenLORIS-Object dataset. The results demonstrate that the object recognition task in the ever-changing difficulty environments is far from being solved and the bottlenecks are at the forward/backward transfer designs. Our dataset and benchmark are publicly available at at \href{https://lifelong-robotic-vision.github.io/dataset/object}{\underline{https://lifelong-robotic-vision.github.io/dataset/object}}.

Eoin Brophy, Zhengwei Wang, Tomas E. Ward

In the recent years Generative Adversarial Networks (GANs) have demonstrated significant progress in generating authentic looking data. In this work we introduce our simple method to exploit the advancements in well established image-based GANs to synthesise single channel time series data. We implement Wasserstein GANs (WGANs) with gradient penalty due to their stability in training to synthesise three different types of data; sinusoidal data, photoplethysmograph (PPG) data and electrocardiograph (ECG) data. The length of the returned time series data is limited only by the image resolution, we use an image size of 64x64 pixels which yields 4096 data points. We present both visual and quantitative evidence that our novel method can successfully generate time series data using image-based GANs.

Zhengwei Wang, Graham Healy, Alan F. Smeaton et al.

Rapid Serial Visual Presentation (RSVP) is a paradigm that supports the application of cortically coupled computer vision to rapid image search. In RSVP, images are presented to participants in a rapid serial sequence which can evoke Event-related Potentials (ERPs) detectable in their Electroencephalogram (EEG). The contemporary approach to this problem involves supervised spatial filtering techniques which are applied for the purposes of enhancing the discriminative information in the EEG data. In this paper we make two primary contributions to that field: 1) We propose a novel spatial filtering method which we call the Multiple Time Window LDA Beamformer (MTWLB) method; 2) we provide a comprehensive comparison of nine spatial filtering pipelines using three spatial filtering schemes namely, MTWLB, xDAWN, Common Spatial Pattern (CSP) and three linear classification methods Linear Discriminant Analysis (LDA), Bayesian Linear Regression (BLR) and Logistic Regression (LR). Three pipelines without spatial filtering are used as baseline comparison. The Area Under Curve (AUC) is used as an evaluation metric in this paper. The results reveal that MTWLB and xDAWN spatial filtering techniques enhance the classification performance of the pipeline but CSP does not. The results also support the conclusion that LR can be effective for RSVP based BCI if discriminative features are available.

Eoin Brophy, José Juan Dominguez Veiga, Zhengwei Wang et al.

The current gold standard for human activity recognition (HAR) is based on the use of cameras. However, the poor scalability of camera systems renders them impractical in pursuit of the goal of wider adoption of HAR in mobile computing contexts. Consequently, researchers instead rely on wearable sensors and in particular inertial sensors. A particularly prevalent wearable is the smart watch which due to its integrated inertial and optical sensing capabilities holds great potential for realising better HAR in a non-obtrusive way. This paper seeks to simplify the wearable approach to HAR through determining if the wrist-mounted optical sensor alone typically found in a smartwatch or similar device can be used as a useful source of data for activity recognition. The approach has the potential to eliminate the need for the inertial sensing element which would in turn reduce the cost of and complexity of smartwatches and fitness trackers. This could potentially commoditise the hardware requirements for HAR while retaining the functionality of both heart rate monitoring and activity capture all from a single optical sensor. Our approach relies on the adoption of machine vision for activity recognition based on suitably scaled plots of the optical signals. We take this approach so as to produce classifications that are easily explainable and interpretable by non-technical users. More specifically, images of photoplethysmography signal time series are used to retrain the penultimate layer of a convolutional neural network which has initially been trained on the ImageNet database. We then use the 2048 dimensional features from the penultimate layer as input to a support vector machine. Results from the experiment yielded an average classification accuracy of 92.3%. This result outperforms that of an optical and inertial sensor combined (78%) and illustrates the capability of HAR systems using...

Zhengwei Wang, Graham Healy, Alan F. Smeaton et al.

There is a growing interest in using generative adversarial networks (GANs) to produce image content that is indistinguishable from real images as judged by a typical person. A number of GAN variants for this purpose have been proposed, however, evaluating GANs performance is inherently difficult because current methods for measuring the quality of their output are not always consistent with what a human perceives. We propose a novel approach that combines a brain-computer interface (BCI) with GANs to generate a measure we call Neuroscore, which closely mirrors the behavioral ground truth measured from participants tasked with discerning real from synthetic images. This technique we call a neuro-AI interface, as it provides an interface between a human's neural systems and an AI process. In this paper, we first compare the three most widely used metrics in the literature for evaluating GANs in terms of visual quality and compare their outputs with human judgments. Secondly we propose and demonstrate a novel approach using neural signals and rapid serial visual presentation (RSVP) that directly measures a human perceptual response to facial production quality, independent of a behavioral response measurement. The correlation between our proposed Neuroscore and human perceptual judgments has Pearson correlation statistics: $\mathrm{r}(48) = -0.767, \mathrm{p} = 2.089e-10$. We also present the bootstrap result for the correlation i.e., $\mathrm{p}\leq 0.0001$. Results show that our Neuroscore is more consistent with human judgment compared to the conventional metrics we evaluated. We conclude that neural signals have potential applications for high quality, rapid evaluation of GANs in the context of visual image synthesis.