Subsequent Task Research Articles

Image Deraining Algorithm Based on Multi-Scale Features

In target detection, tracking, and recognition tasks, high-quality images can achieve better results. However, in actual scenarios, the visual effects and data quality of images are greatly reduced due to the influence of environmental factors, which affect subsequent detection, recognition, and other tasks. Therefore, this paper proposes an image rain removal algorithm based on multi-scale features, which can effectively remove rain streaks. First of all, this paper proposes a deraining algorithm that combines spatial information to improve the network’s generalization ability on real images, aiming at the problem of synthetic datasets used by previous deraining algorithms. Then, by proposing a multi-scale rain removal algorithm, it improves the feature extraction capabilities of existing deraining algorithms. Before extracting deep rain features, a preliminary fusion of multi-scale shallow features can be performed, which can show better performance in images of different sizes. In addition, a spatial attention module and channel are introduced. The attention module increases the ability to extract rain information at each scale; the resulting multi-scale feature image rain removal algorithm is called MFD. Finally, the rain removal algorithm is validated on the rain removal dataset, and the proposed method can effectively remove rain patterns, provide strong performance improvement on several datasets in the image rain removal task, and provide high-quality images for subsequent detection and recognition tasks.

Attention switching through text dissimilarity: a cognition research on fragmented reading behavior.

People tend to obtain information through fragmented reading. However, this behavior itself might lead to distraction and affect cognitive ability. To address it, it is necessary to understand how fragmented reading behavior influences readers' attention switching. In this study, the researchers first collected online news that had 6 theme words and 60 sentences to compose the experimental material, then defined the degree of text dissimilarity, used to measure the degree of attention switching based on the differences in text content, and conducted an EEG experiment based on P200. The results showed that even after reading the fragmented text content with the same overall content, people in subsequent cognitive tasks had more working memory capacity, lower working memory load, and less negative impact on cognitive ability with the text content with lower text dissimilarity. Additionally, attention switching caused by differences in concept or working memory representation of text content might be the key factor affecting cognitive ability in fragmented reading behavior. The findings disclosed the relation between cognitive ability and fragmented reading and attention switching, opening a new perspective on the method of text dissimilarity. This study provides some references on how to reduce the negative impact of fragmented reading on cognitive ability on new media platforms.

Effect of additional tasks on the reaction time of braking responses in simulated car driving: beyond the PRP effect.

The presentation of one task increases the reaction time on a subsequent task, if stimulus onset asynchrony (SOA) between tasks is short. This psychological refractory period (PRP) effect is typically leveling off as SOA approaches 1 s, which has been documented both in classical laboratory paradigms and in simulated car driving. Here we report a more persistent effect on the subsequent task that goes well beyond the typical duration of the PRP effect. In a driving simulator, 120 healthy older participants followed a lead car that mostly drove at a constant speed. They had to maintain a regular distance from the lead car and had to brake when the lead car braked. Participants also engaged in several additional tasks during driving (two types of tasks: typing three-digit numbers, stating arguments on public issues). SOA between the braking task and the last preceding additional task was 11.49 s ± 1.99 (mean and standard deviation). In a control condition, the braking task was administered without additional tasks. Main performance outcome was Braking Reaction Time (RT, in s), as the interval between onset of brake lights of the lead car and the moment participants released the gas pedal. Additionally, foot movement time (MT, in s), i.e., the difference between gas pedal release and brake pedal onset, was considered for possible compensation behavior. Inter-vehicle distance to the lead car (in m) was taken into account as a moderator. We found that RT averaged 0.77 s without additional tasks, but averaged 1.45 s with additional tasks. This RT difference was less pronounced at smaller inter-vehicle distances, and was not compensated for by faster MT from the gas pedal to the brake pedal. We conclude that detrimental effects of additional tasks on subsequent braking responses can be more persistent than suggested by the PRP effect, possibly because of maintaining multiple task sets, requiring increased executive control. We further conclude that potential detrimental effects can be ameliorated at small inter-vehicle distances by mobilizing extra cognitive resources when response urgency is higher. As a practical implication of our study, distracting stimuli can have persisting detrimental effects on traffic safety.

What makes a stimulus worthy of attention: Cue-outcome correlation and choice relevance in the learned predictiveness effect.

The learned predictiveness effect refers to the tendency for predictive cues to attract greater attention and show faster learning in subsequent tasks. However, in typical designs, the predictiveness of each cue (its objective cue-outcome correlation) is confounded with the degree to which it is informative for making the correct response on each trial (a feature we term choice relevance). In four experiments, we tested the unique contributions of cue-outcome correlation and choice relevance to the learned predictiveness effect by manipulating the outcome choices available on each trial. Experiments 1A and 1B compared two sets of partially predictive cues and found that participants learned more in a transfer phase about the set of cues that were previously choice-relevant. Experiments 2A and 2B used a design in which the cue-outcome correlation was stronger for one set of cues (perfect predictors) than the other set (imperfect predictors). Manipulating the choice relevance of the imperfect predictors in this design did not influence the magnitude of the learning bias toward the perfect predictor. Unlike cue-outcome correlation, choice relevance did not seem to correspond to biases in eye-gaze, suggesting that it operates via a distinct mechanism. Simulations with a modified EXIT model successfully predicted cue-outcome correlation and choice relevance effects by assuming that participants update learning for present outcomes only, but incorrectly predicted additive effects. We conclude that cue-outcome correlation and choice relevance are important factors that can lead to biases in future learning; both were individually sufficient but neither was necessary. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Aberration aware feature detection and description

Feature detection and description are crucial for image matching, as better performance at this stage leads to more accurate matching results, which is essential for subsequent vision-based tasks. However, images captured by different optical systems may suffer from various optical aberrations, especially in off-axis field and out-of-depth-of-field regions, making it challenging for models to extract consistent feature locations and descriptors. In this paper, we propose what we believe to be a novel method for training feature detection and description networks by incorporating optical system aberrations modeled by point spread function(PSF). We introduce PSF augmentation and homographic PSF adaptation, which customize the training of feature detection and description models for specific optical systems using general unlabeled image datasets. Experimental results demonstrate that our method significantly improves the performance of feature detection and description in images captured by given cameras.

Unsupervised transfer learning for fault diagnosis across similar chemical processes

Fault diagnosis plays a crucial role in chemical processes to prevent major accidents. Recent advancements have leveraged deep learning to enhance fault diagnosis capabilities significantly. However, the success of deep learning diagnosis models primarily relies on access to the extensive labeled dataset of faults. Another limitation is the variation of ambient temperature and raw material components contributing to diverse operating conditions. These challenges lead to the current fault diagnosis techniques being unsuitable for universal application across different devices. In this paper, a novel unsupervised transfer learning method called multiple source domain adaptation network (MSDAN) is proposed for industrial fault diagnosis across similar chemical processes. Owing to the limited availability of fault data in industrial manufacturing, fault samples for model training are expanded via an efficient generative adversarial network variant. Categorical features of various source domains are precisely extracted through distinct channels by an integration model of Transformer and multiscale convolutional neural network. Multi-channel Domain adaptation based on polynomial kernel-induced maximum mean discrepancy aligns the joint distributions of multiple domains and facilitates subsequent classification tasks. The effectiveness and robustness of the proposed method are demonstrated through experiments conducted on the multimode Tennessee Eastman process and the real-world fluid catalytic cracking process across different units.

Discovering the signal subgraph: An iterative screening approach on graphs

Supervised learning on graphs is a challenging task due to the high dimensionality and inherent structural dependencies in the data, where each edge depends on a pair of vertices. Existing conventional methods are designed for standard Euclidean data and do not account for the structural information inherent in graphs. In this paper, we propose an iterative vertex screening method to achieve dimension reduction across multiple graph datasets with matched vertex sets and associated graph attributes. Our method aims to identify a signal subgraph to provide a more concise representation of the full graphs, potentially benefiting subsequent vertex classification tasks. The method screens the rows and columns of the adjacency matrix concurrently and stops when the resulting distance correlation is maximized. We establish the theoretical foundation of our method by proving that it estimates the true signal subgraph with high probability. Additionally, we establish the convergence rate of classification error under the Erdos-Renyi random graph model and prove that the subsequent classification can be asymptotically optimal, outperforming the entire graph under high-dimensional conditions. Our method is evaluated on various simulated datasets and real-world human and murine graphs derived from functional and structural magnetic resonance images. The results demonstrate its excellent performance in estimating the ground-truth signal subgraph and achieving superior classification accuracy.

Risk analysis and assessment of underwater glider mission failure using Dynamic Bayesian Network method

The Autonomous Underwater Glider (AUG), with its extended operational endurance, encounters various uncertain risks during diverse mission tasks. Therefore, conducting dynamic risk analysis is crucial for implementing subsequent decisions and improving equipment safety. This paper combines Fault Tree Analysis (FTA) and Dynamic Bayesian Network (DBN) to develop a customized dynamic risk assessment method for AUG task failure under multiple risk factors. Firstly, preliminary risk factors are identified by FTA using existing experimental data. Next, transformation rules are constructed to map FTA into a basic DBN model. Then, parameters such as conditional probabilities are determined through logic gates and parameter learning, and the proposed model is validated based on axioms. Finally, key risk factors are identified based on actual failure data. The research results indicate that the key risk factors leading to AUG task failure are progressive abnormalities such as biofouling and ocean current disturbance. The method proposed in this paper can achieve dynamic risk assessment, providing strong references for operators in making decisions for subsequent tasks.

Ego depletion does not negatively impact novice sprinters’ start performance during a timed 20m all-out sprint

Temporary reductions in self-control due to previous cognitive effort, also known as ‘ego depletion’, have been shown to negatively impact performance in subsequent tasks requiring self-control. Various theoretical accounts suggest that high levels of task motivation may prevent negative behavioural effects of ego depletion. Against this background, the current study performed a conceptual replication of a previous study examining the effects of ego depletion on sprint start performance (i.e., Englert et al., 2015) and implemented additional monetary incentives to manipulate task motivation. Using a pretest-intervention-posttest design, ninety-six novice sprinters performed a sprinting task that required them to complete a series of 20-m sprints. In between pretest and posttest, participants performed a letter writing task either under high cognitive demand (depletion group) or low cognitive demand (non-depletion control group). Subsequently, and prior to starting the posttest, participants were presented either with a monetary incentive or no incentive. Manipulation checks confirmed that the letter writing task successfully manipulated perceived regulatory demand and that task motivation at the posttest was higher for participants who were offered a monetary incentive. The number of false starts decreased significantly with ego depletion, while sprint start reaction times and sprint finish times remained unaffected. Regardless of ego depletion, sprint finish times significantly improved with monetary incentives. These results suggest that brief engagement in a demanding cognitive task is unlikely to impair novice sprinters’ ability to self-regulate during the start of a subsequent, maximal sprinting effort for which they are highly motivated.

Fatigued or bored? Investigating the effect of different types of mental fatigue on 3 km running performance

Mental fatigue has been highly cited as having a negative impact on endurance performance. Few, however, have investigated whether different types of mental fatigue, namely active and passive fatigue, might affect endurance performance differently. This study used a repeated-measures design where 11 participants completed a 3 km run after three 32-min conditions: an actively fatiguing task (Task-load Dual-back; TloadDback); a passively fatiguing task (Mackworth Clock); and a control task (Documentary). Subjective ratings and performance on a second task (Flanker task) were taken before and after the 32-min tasks, while ratings of perceived effort, motivation and workload were taken during the 3 km run. Results showed that both fatigue conditions were mentally fatiguing, with the TloadDback rated as more demanding and the Mackworth Clock more boring and less motivating. Performance on subsequent tasks showed different effects: the TloadDback condition had the slowest responses on the post-test Flanker task, while the Mackworth Clock condition had the slowest completion time on the 3 km run, though this difference in completion times was non-significant. These results suggest that different cognitive tasks lead to different types of mental fatigue, which can lead to different outcomes on subsequent cognitive tasks but non-significant differences on subsequent physical tasks. This study highlights that to understand the influence of mental fatigue on physical performance, it is important to understand the cognitive tasks used and how different cognitive and physical tasks interact. Future work should examine whether performance outcomes in other areas commonly affected by mental fatigue, like technical and tactical sporting skills, are affected differently by active and passive fatigue.

Self-supervised learning for classifying paranasal anomalies in the maxillary sinus

PurposeParanasal anomalies, frequently identified in routine radiological screenings, exhibit diverse morphological characteristics. Due to the diversity of anomalies, supervised learning methods require large labelled dataset exhibiting diverse anomaly morphology. Self-supervised learning (SSL) can be used to learn representations from unlabelled data. However, there are no SSL methods designed for the downstream task of classifying paranasal anomalies in the maxillary sinus (MS).MethodsOur approach uses a 3D convolutional autoencoder (CAE) trained in an unsupervised anomaly detection (UAD) framework. Initially, we train the 3D CAE to reduce reconstruction errors when reconstructing normal maxillary sinus (MS) image. Then, this CAE is applied to an unlabelled dataset to generate coarse anomaly locations by creating residual MS images. Following this, a 3D convolutional neural network (CNN) reconstructs these residual images, which forms our SSL task. Lastly, we fine-tune the encoder part of the 3D CNN on a labelled dataset of normal and anomalous MS images.ResultsThe proposed SSL technique exhibits superior performance compared to existing generic self-supervised methods, especially in scenarios with limited annotated data. When trained on just 10% of the annotated dataset, our method achieves an area under the precision-recall curve (AUPRC) of 0.79 for the downstream classification task. This performance surpasses other methods, with BYOL attaining an AUPRC of 0.75, SimSiam at 0.74, SimCLR at 0.73 and masked autoencoding using SparK at 0.75.ConclusionA self-supervised learning approach that inherently focuses on localizing paranasal anomalies proves to be advantageous, particularly when the subsequent task involves differentiating normal from anomalous maxillary sinuses. Access our code at https://github.com/mtec-tuhh/self-supervised-paranasal-anomaly.

Enhanced image steganalysis through reinforcement learning and generative adversarial networks

With the surging prominence of digital communication platforms, there has been an intensified emphasis on ensuring robust security and privacy measures. Against this backdrop, image steganalysis has emerged as a critical discipline, employing advanced methods to detect clandestine data within image files. At the core of our research is an innovative exploration into image steganalysis using an amalgamation of enhanced reinforcement learning techniques and online data augmentation. This methodology ensures the meticulous identification of concealed data within images. Our design integrates triple parallel dilated convolutions, enabling concurrent extraction of feature vectors from the input images. Once extracted, these vectors are synthesized, paving the way for subsequent classification tasks. To substantiate the efficacy of our approach, we conducted tests on a comprehensive dataset sourced from BossBase 1.01. Furthermore, to discern the influence of transfer learning on our proposed model, the BOWS dataset was employed. Notably, these datasets present a challenge due to its inherent imbalance. To counteract this, we incorporated an advanced Reinforcement Learning (RL) framework. Herein, the dataset samples are envisioned as states in a sequence of interrelated decisions, with the neural network playing the role of the decision-making agent. This agent is then incentivized or reprimanded based on its accuracy in discerning between the minority and majority classes. To bolster our classification capabilities, we innovatively employed data augmentation using images generated by a Generative Adversarial Network (GAN). Concurrently, a regularization mechanism was instituted to alleviate prevalent GAN-related challenges, such as mode collapse and unstable training dynamics. Our experimental outcomes underscore the potency of our methodology. The results highlight a remarkable capability to discern between pristine and steganographic images, registering an average accuracy rate of 85%.

An Efficient Transformer–CNN Network for Document Image Binarization

Color image binarization plays a pivotal role in image preprocessing work and significantly impacts subsequent tasks, particularly for text recognition. This paper concentrates on document image binarization (DIB), which aims to separate an image into a foreground (text) and background (non-text content). We thoroughly analyze conventional and deep-learning-based approaches and conclude that prevailing DIB methods leverage deep learning technology. Furthermore, we explore the receptive fields of pre- and post-network training to underscore the Transformer model’s advantages. Subsequently, we introduce a lightweight model based on the U-Net structure and enhanced with the MobileViT module to capture global information features in document images better. Given its adeptness at learning both local and global features, our proposed model demonstrates competitive performance on two standard datasets (DIBCO2012 and DIBCO2017) and good robustness on the DIBCO2019 dataset. Notably, our proposed method presents a straightforward end-to-end model devoid of additional image preprocessing or post-processing, eschewing the use of ensemble models. Moreover, its parameter count is less than one-eighth of the model, which achieves the best results on most DIBCO datasets. Finally, two sets of ablation experiments are conducted to verify the effectiveness of the proposed binarization model.

MCDHGN: heterogeneous network-based cancer driver gene prediction and interpretability analysis.

Accurately predicting the driver genes of cancer is of great significance for carcinogenesis progress research and cancer treatment. In recent years, more and more deep-learning-based methods have been used for predicting cancer driver genes. However, deep-learning algorithms often have black box properties and cannot interpret the output results. Here, we propose a novel cancer driver gene mining method based on heterogeneous network meta-paths (MCDHGN), which uses meta-path aggregation to enhance the interpretability of predictions. MCDHGN constructs a heterogeneous network by using several types of multi-omics data that are biologically linked to genes. And the differential probabilities of SNV, DNA methylation, and gene expression data between cancerous tissues and normal tissues are extracted as initial features of genes. Nine meta-paths are manually selected, and the representation vectors obtained by aggregating information within and across meta-path nodes are used as new features for subsequent classification and prediction tasks. By comparing with eight homogeneous and heterogeneous network models on two pan-cancer datasets, MCDHGN has better performance on AUC and AUPR values. Additionally, MCDHGN provides interpretability of predicted cancer driver genes through the varying weights of biologically meaningful meta-paths. https://github.com/1160300611/MCDHGN.

Detecting sun glint in UAV RGB images at different times using a deep learning algorithm

Unmanned aerial vehicle (UAV) remote sensing has played a crucial role in water quality monitoring. However, the presence of water sun glint, resulting from specular reflection on the water surface, poses an inevitable challenge in UAV-acquired images. These excessively bright pixels disrupt the original images' spectral and textural characteristics, significantly diminishing their usability. This disruption has repercussions on subsequent tasks, such as target object classification and water quality parameter inversion. Precise detection of sun glint is a prerequisite for removing them, but current methods suffer from missed and false detections. In this study, we collected images by UAV to construct a specialized dataset for water surface sun glint, namely water sun glint detection (WSGD) dataset, laying the groundwork for further research endeavors. We proposed the Res_AUNet network by enhancing the UNet convolutional neural network. The Convolutional Block Attention Module was integrated into the encoding-decoding skip connections of the network, we also refined the convolutional blocks to better capture the distinctive semantic features associated with water sun glint. To mitigate overfitting, the residual structures were incorporated and the number of convolutional kernels within each block was reduced. The Res_AUNet network was trained and evaluated using the WSGD dataset, achieving metrics with an Accuracy of 98.02%, an F1-score of 83.67%, and an IOU of 74.73%. These results underscore the precision of our proposed method for water sun glint detection in UAV water images, offering valuable insights for effectively eliminating water sun glint and determining the optimal timing for UAV water image acquisition.

High-precision air conditioning load forecasting model based on improved sparrow search algorithm

Air conditioning load prediction is a crucial prerequisite for ensuring the efficient operation of air conditioning systems and enhancing demand-side response in electricity. Due to the uncertainty in model parameter settings and limitations of the Sparrow Search Algorithm (SSA2) itself, leading to significant human and material resources consumption during the prediction process, this study improves SSA (ISSA) to enhance air conditioning prediction efficiency and accuracy. The enhanced SSA is then combined with Long Short-Term Memory (LSTM). Firstly, Singular Spectrum Analysis (SSA1) is employed for feature extraction on the load to obtain more regular sub-signals, facilitating subsequent prediction tasks. Simultaneously, the Maximum Information Coefficient (MIC) is used to measure the correlation between data, assisting in selecting input variables for the model. Subsequently, each sub-signal is input into the ISSA-LSTM model for prediction, and the prediction results of each sub-signal are reconstructed to obtain the final prediction. Finally, data from air conditioning loads in two different types of buildings are collected to validate the model. Upon validation, it was found that the SSA1-ISSA- LSTM model exhibits superior performance compared to other advanced combination models. When predicting air conditioning loads in office building, the R2 value reaches 0.9955, with MAPE, MAE, and RMSE being 0.73 %, 4.658RT, and 6.1151RT respectively. When predicting air conditioning loads in medical building, the R2 is as high as 0.9920, with MAPE, MAE, and RMSE being 0.55 %, 5.8543RT, and 7.9599RT respectively.

Thermal image visualization using multi-discriminator CycleGAN with unpaired thermal–visible image training set

Thermal imaging utilizes the differences in thermal radiation within a scene to capture images, effectively compensating for the limited imaging capabilities of visible-light cameras in low-light environments. In order to further enhance thermal imaging, we propose a novel approach that transforms the visualization of thermal images into an image conversion task. Our method, named TIVNet (Thermal Images Visualization Network), is based on a multi-discriminator cycle-consistent adversarial network (CycleGAN) and operates in the YUV color space to accurately colorize and represent thermal infrared images. The utilization of color and texture discriminators in our framework ensures that the converted images exhibit appropriate color styles and capture fine details, respectively. In the generator, we employed a Dense Residual Module (DRM) to facilitate the reuse of features and enhance the flow of information. By utilizing the DRM, we aim to enhance the overall performance and efficiency of the generator network. Additionally, we incorporated the Dual-channel Self-Attention Module(DCSA) to enhance the feature representation. By leveraging self-attention, the generator can effectively learn and exploit the inherent relationships between different regions, leading to improved feature representation and, consequently, more accurate colorization and representation of thermal infrared images. Notably, our proposed method leverages unsupervised learning, eliminating the need for a fully paired thermal infrared-visible image dataset. Through extensive experimentation on publicly available datasets as well as our own, we demonstrate that our approach outperforms state-of-the-art methods in thermal image visualization. Furthermore, the converted images produced by TIVNet exhibit a high level of realism, thereby enhancing the visual impact of thermal imaging. This improvement is particularly beneficial for human perception and subsequent target recognition tasks. Our findings suggest that the application of our method yields superior results both quantitatively and qualitatively, showcasing its potential for advancing thermal imaging capabilities.

Deblurring Network for UV Images of Colorless Chemicals Floating on Sea Surface

Automatic imaging monitoring technology has been developed as a mature solution to address oil spill accidents. However, when airborne equipment is used to monitor the area of leakage, there is a chance of relative motion between the equipment and the target, leading to blurred images. This can hinder subsequent monitoring tasks, such as the identification and segmentation of hazardous chemical targets. Therefore, it is crucial to deblur these images, as sharper images can substantially improve the precision and accuracy of these tasks. In light of this, this paper presents a novel multi-scale recurrent deblurring network, leveraging the Convolutional Block Attention Module (CBAM) attention mechanism, to restore clear images in an end-to-end manner. The CBAM attention mechanism has been incorporated, which learns features in both the spatial and channel domains, thereby enhancing the deblurring ability of the network by combining the attention mechanisms of multiple domains. To evaluate the effectiveness of the proposed method, we applied the deblurring network to our self-built dataset of blurred ultraviolet (UV) images of colorless chemicals floating on water surfaces. Compared to the SRN-Deblur deblurring network, the proposed approach yields improved PSNR and SSIM results. Moreover, when compared with the state-of-the-art DeblurGANv2 method, it performs better in the SSIM. These results demonstrate that the proposed multi-scale recurrent deblurring network, based on the CBAM attention mechanism, exhibits a superior deblurring performance.

The effect of false cognitive feedback on subsequent cognitive task performance

ABSTRACT Introduction Previous research has found beliefs about oneself and one’s own abilities may have the potential to affect subsequent performance on a particular task. Additionally, providing false feedback about a particular characteristic or even about overall cognitive abilities may also affect performance on later tasks. However, it is unclear to what extent false positive or negative feedback about cognition will affect subsequent executive function task performance. In the present series of studies, we examined whether receiving negative false feedback about cognition would affect subsequent decision making and other executive function task performance. Method In Study 1, the participants (n = 115) received false feedback that they were either high or low in creative intelligence before completing a series of decision making tasks. In Study 2, the participants (n = 146) completed a similar false feedback paradigm before completing assessments of a range of executive functions. Results Across studies, we found limited evidence of a consistent pattern of how false feedback affects subsequent cognitive task performance, although receiving positive and negative feedback affected specific tasks. Conclusions Our results indicate that the influence of false feedback on task performance is variable and may depend on factors such as the specific task or executive function assessed. In clinical work, it is important to consider how patients may internalize feedback about their cognitive abilities, as the feedback, coupled with other factors such as level of insight, apathy, disinhibition, or prior perceptions regarding a diagnosis, may influence interpretations.

A multi-resolution self-supervised learning framework for semantic segmentation in histopathology

Modern whole slide imaging technique together with supervised deep learning approaches have been advancing the field of histopathology, enabling accurate analysis of tissues. These approaches use whole slide images (WSIs) at various resolutions, utilising low-resolution WSIs to identify regions of interest in the tissue and high-resolution for detailed analysis of cellular structures. Due to the labour-intensive process of annotating gigapixels WSIs, accurate analysis of WSIs remains challenging for supervised approaches. Self-supervised learning (SSL) has emerged as an approach to build efficient and robust models using unlabelled data. It has been successfully used to pre-train models to learn meaningful image features which are then fine-tuned with downstream tasks for improved performance compared to training models from scratch. Yet, existing SSL methods optimised for WSI are unable to leverage the multi-resolutions and instead, work only in an individual resolution neglecting the hierarchical structure of multi-resolution inputs. This limitation prevents from the effective utilisation of complementary information between different resolutions, hampering discriminative WSI representation learning. In this paper we propose a Multi-resolution SSL Framework for WSI semantic segmentation (MSF-WSI) that effectively learns histopathological features. Our MSF-WSI learns complementary information from multiple WSI resolutions during the pre-training stage; this contrasts with existing works that only learn between the resolutions at the fine-tuning stage. Our pre-training initialises the model with a comprehensive understanding of multi-resolution features which can lead to improved performance in the subsequent tasks. To achieve this, we introduced a novel Context-Target Fusion Module (CTFM) and a masked jigsaw pretext task to facilitate the learning of multi-resolution features. Additionally, we designed Dense SimSiam Learning (DSL) strategy to maximise the similarities of image features from early model layers to enable discriminative learned representations. We evaluated our method using three public datasets on breast and liver cancer segmentation tasks. Our experiment results demonstrated that our MSF-WSI surpassed the accuracy of other state-of-the-art methods in downstream fine-tuning and semi-supervised settings.