The human vision and perception system is inherently incremental where new knowledge is continual... more The human vision and perception system is inherently incremental where new knowledge is continually learned over time whilst existing knowledge is retained. On the other hand, deep learning networks are ill-equipped for incremental learning. When a well-trained network is adapted to new categories, its performance on the old categories will dramatically degrade. To address this problem, incremental learning methods have been explored which preserve the old knowledge of deep learning models. However, the state-of-the-art incremental object detector employs an external fixed region proposal method that increases overall computation time and reduces accuracy comparing to Region Proposal Network (RPN) based object detectors such as Faster RCNN. The purpose of this paper is to design an efficient end-to-end incremental object detector using knowledge distillation. We first evaluate and analyze the performance of the RPN-based detector with classic distillation on incremental detection tasks. Then, we introduce multi-network adaptive distillation that properly retains knowledge from the old categories when fine-tuning the model for new task. Experiments on the benchmark datasets, PASCAL VOC and COCO, demonstrate that the proposed incremental detector based on Faster RCNN is more accurate as well as being 13 times faster than the baseline detector.
Despite much progress being made in the field of object recognition with the advances of deep lea... more Despite much progress being made in the field of object recognition with the advances of deep learning, there are still several factors negatively affecting the performance of deep learning models. Domain shift is one of these factors and is caused by discrepancies in the distributions of the testing and training data. In this paper, we focus on the problem of compact feature clustering in domain generalization to help optimize the embedding space from multi-domain data. We design a domainaware triplet loss for domain generalization to help the model to not only cluster similar semantic features, but also to disperse features arising from the domain. Unlike previous methods focusing on distribution alignment, our algorithm is designed to disperse domain information in the embedding space. The basic idea is motivated based on the assumption that embedding features can be clustered based on domain information, which is mathematically and empirically supported in this paper. In addition, during our exploration of feature clustering in domain generalization, we note that factors affecting the convergence of metric learning loss in domain generalization are more important than the pre-defined domains. To solve this issue, we utilize two methods to normalize the embedding space, reducing the internal covariate shift of the embedding features. The ablation study demonstrates the effectiveness of our algorithm. Moreover, the experiments on the benchmark datasets, including PACS, VLCS and Office-Home, show that our method outperforms related methods focusing on domain discrepancy. In particular, our results on RegnetY-16 are significantly better than state-of-the-art methods on the benchmark datasets. Our code will be released at https://github.com/workerbcd/DCT.
We study the newly formed (with an upper limit of ∼101 min) subauroral flows in the topside ionos... more We study the newly formed (with an upper limit of ∼101 min) subauroral flows in the topside ionosphere. These include the dawnside subauroral flows observed at ∼5–6 magnetic local time (MLT), the duskside sub‐auroral polarization streams (SAPS) at ∼15 MLT, and the nightside sub‐auroral ion drifts (SAID) at ∼19–20 MLT. We employ the same types of observations for investigating the various underlying magnetospheric processes related to the development of the newly formed subauroral flows. These processes include (a) the stretching of magnetic field lines (leading to plasma density dropouts) due to the unfolding magnetotail reconnection, (b) the dawnside and nightside diversions of the magnetotail‐reconnection—related earthward plasma flows, (c) the ongoing particle injections, and (d) the trapping of ring current protons at subauroral and auroral latitudes. Based on these processes (a–d), we found strong similarities in the development of the newly formed dawnside and duskside/nightsi...
The continual appearance of new objects in the visual world poses considerable challenges for cur... more The continual appearance of new objects in the visual world poses considerable challenges for current deep learning methods in realworld deployments. The challenge of new task learning is often exacerbated by the scarcity of data for the new categories due to rarity or cost. Here we explore the important task of Few-Shot Class-Incremental Learning (FSCIL) and its extreme data scarcity condition of one-shot. An ideal FSCIL model needs to perform well on all classes, regardless of their presentation order or paucity of data. It also needs to be robust to open-set real-world conditions and be easily adapted to the new tasks that always arise in the field. In this paper, we first reevaluate the current task setting and propose a more comprehensive and practical setting for the FSCIL task. Then, inspired by the similarity of the goals for FSCIL and modern face recognition systems, we propose our method-Augmented Angular Loss Incremental Classification or ALICE. In ALICE, instead of the commonly used cross-entropy loss, we propose to use the angular penalty loss to obtain well-clustered features. As the obtained features not only need to be compactly clustered but also diverse enough to maintain generalization for future incremental classes, we further discuss how class augmentation, data augmentation, and data balancing affect classification performance. Experiments on benchmark datasets, including CIFAR100, miniImageNet, and CUB200, demonstrate the improved performance of ALICE over the state-of-the-art FSCIL methods.
2019 Digital Image Computing: Techniques and Applications (DICTA)
Galvanised steel transmission towers in electrical power grids suffer from corrosion to different... more Galvanised steel transmission towers in electrical power grids suffer from corrosion to different levels depending on age and environment. To ensure the power grid can operate safely, significant resources are spent to monitor the corrosion level of towers. Photographs from helicopters, drones, and a variety of staffs are often used to capture condition, however, these images still need manual inspection to determine the corrosion level before carrying out maintenance works. In this paper, we describe a framework employing multiple deep neural networks based classifiers and detectors to perform automatic image-based condition monitoring for steel transmission towers. Given a random variety of images of a tower, our proposed framework will first determine the location of the image on the structure via a trained zone classifier. Then, fine-grain corrosion inspection will be performed on both fasteners and structural members, respectively. In addition, an automatic zoomin functionality will be applied to images which have high resolution but are a long distance away. This step will ensure the detection performance on small objects on the tower. Finally, the overall corrosion status report for this tower will be calculated and generated automatically. Additionally, we released a subset of our data to contribute to this novel direction. Experiments show that our framework can assess the tower efficiently.
2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2017
Gram stained direct smears test is clinically useful in early identification of infections. Unfor... more Gram stained direct smears test is clinically useful in early identification of infections. Unfortunately, this practice is considered time consuming and labour intensive. Most existing effort in this area is to perform highmagnification analysis of images taken from manually selected areas. In this paper, we address the problem of the automatic selection of candidate areas (or patches) for subsequent high-magnification analysis. Specifically, we explore and study the possibility of selecting good working areas based on low-magnification images, where bacteria are likely to be found when viewed in high-magnification images. To this end, we develop an approach to classify the areas of interest according to the textural information of the image patch. We explore and study the efficacy of traditional textural features such as Histogram of Gradients, Local Binary Patterns, and 2 Dimensional Discrete Cosine Transform. Experiments show that the best variant method is able to select working areas where it is likely to find bacteria in high-powered objective images in a wide range of images.
Pathology tests play a major role in our health care system and they assist physicians to make ac... more Pathology tests play a major role in our health care system and they assist physicians to make accurate diagnoses. Recently, there has been growing interest in developing Computer-Aided Diagnostic (CAD) systems which should further improve the reliability and consistency of these tests. We discuss the Anti-Nuclear Antibodies (ANA) test via Indirect Immuno uorescence protocol on Human Epithelial Type 2 (HEp- 2) cells which is normally used to determine the presence of Connective Tissue Diseases such as Sjffogren's syndrome, Systemic Lupus Erythematosus and Rheumatoid Arthritis. One task in the test is to determine the pattern of positive HEp-2 cells. This chapter discusses and compares both traditional and contemporary approaches to this problem via image-based analysis. Whilst, the traditional approaches comprise several popular methods used to address generic image classification problems, the contemporary approaches apply advanced mathematical techniques to improve system accu...
2016 23rd International Conference on Pattern Recognition (ICPR), 2016
Automatically recognising facial emotions has drawn increasing attention in computer vision. Faci... more Automatically recognising facial emotions has drawn increasing attention in computer vision. Facial landmark based methods are one of the most widely used approaches to perform this task. However, these approaches do not provide good performance. Thus, researchers usually tend to combine more information such as textural and audio information to increase the recognition rate. In this paper we propose a novel method, here called the landmark manifold, that shows the possibility to achieve competitive performance by facial landmark information alone. Through experiments on the well-known dataset: marked Cohn-Kanade extended facial emotion dataset (CK+), we show that with accurate facial landmarks, our simple approach is fast to run and can achieve competitive performance with enormously expensive methods.
2019 Digital Image Computing: Techniques and Applications (DICTA), 2019
The condition of the Glomeruli, or filter sacks, in renal Direct Immunofluorescence (DIF) specime... more The condition of the Glomeruli, or filter sacks, in renal Direct Immunofluorescence (DIF) specimens is a critical indicator for diagnosing kidney diseases. A digital pathology system which digitizes a glass histology slide into a Whole Slide Image (WSI) and then automatically detects and zooms in on the glomeruli with a higher magnification objective will be extremely helpful for pathologists. In this paper, using glomerulus detection as the study case, we provide analysis and observations on several important issues to help with the development of Computer Aided Diagnostic (CAD) systems to process WSIs. Large image resolution, large file size, and data scarcity are always challenging to deal with. To this end, we first examine image downsampling rates in terms of their effect on detection accuracy. Second, we examine the impact of image compression. Third, we examine the relationship between the size of the training set and detection accuracy. To understand the above issues, experiments are performed on the state-of-the-art detectors: Faster R-CNN, R-FCN, Mask R-CNN and SSD. Critical findings are observed: (1) The best balance between detection accuracy, detection speed and file size is achieved at 8 times downsampling captured with a 40× objective; (2) compression which reduces the file size dramatically, does not necessarily have an adverse effect on overall accuracy; (3) reducing the amount of training data to some extents causes a drop in precision but has a negligible impact on the recall; (4) in most cases, Faster R-CNN achieves the best accuracy in the glomerulus detection task. We show that the image file size of 40× WSI images can be reduced by a factor of over 6000 with negligible loss of glomerulus detection accuracy.
2019 IEEE International Conference on Image Processing (ICIP), 2019
missing texture and edge information. In this paper, we proposed a framework to translate images ... more missing texture and edge information. In this paper, we proposed a framework to translate images while preserving more realistic textures and details. To this end, we disentangle the samples into shared content space and domain-specific style domain. Then, according to the blurred outlines and textures in the source domain, we introduce the classic canny edge detection algorithm to encode the boundary and edge information in the content latent space. We test the proposed method in the thermal to visible image translation scenario and the experimental results demonstrate that the proposed method outperforms several other state-of-the-art models.
2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020
The difficulty of processing gigapixel whole slide images (WSIs) in clinical microscopy has been ... more The difficulty of processing gigapixel whole slide images (WSIs) in clinical microscopy has been a long-standing barrier to implementing computer aided diagnostic systems. Since modern computing resources are unable to perform computations at this extremely large scale, current state of the art methods utilize patch-based processing to preserve the resolution of WSIs. However, these methods are often resource intensive and make significant compromises on processing time. In this paper, we demonstrate that conventional patch-based processing is redundant for certain WSI classification tasks where high resolution is only required in a minority of cases. This reflects what is observed in clinical practice; where a pathologist may screen slides using a low power objective and only switch to a high power in cases where they are uncertain about their findings. To eliminate these redundancies, we propose a method for the selective use of high resolution processing based on the confidence of predictions on downscaled WSIs-we call this the Selective Objective Switch (SOS). Our method is validated on a novel dataset of 684 Liver-Kidney-Stomach immunofluorescence WSIs routinely used in the investigation of autoimmune liver disease. By limiting high resolution processing to cases which cannot be classified confidently at low resolution, we maintain the accuracy of patch-level analysis whilst reducing the inference time by a factor of 7.74.
A method to segment a nonstationary signal into non-overlapping successive near stationary segmen... more A method to segment a nonstationary signal into non-overlapping successive near stationary segments using the Covariance method of Autoregressive (AR) Modelling is presented. An assumption of" local ergodicityn within the near stationary segments allows improved spectral estimation because of the adaptive nature of the windowing_ Performance is evaluated on simulated AR data and real oceanographic signals.
2019 IEEE International Conference on Image Processing (ICIP)
Electrical distribution poles are important assets in electricity supply. These poles need to be ... more Electrical distribution poles are important assets in electricity supply. These poles need to be maintained in good condition to ensure they protect community safety, maintain reliability of supply, and meet legislative obligations. However, maintaining such a large volumes of assets is an expensive and challenging task. To address this, recent approaches utilise imagery data captured from helicopter and/or drone inspections. Whilst reducing the cost for manual inspection, manual analysis on each image is still required. As such, several image-based automated inspection systems have been proposed. In this paper, we target two major challenges: tiny object detection and extremely imbalanced datasets, which currently hinder the wide deployment of the automatic inspection. We propose a novel two-stage zoom-in detection method to gradually focus on the object of interest. To address the imbalanced dataset problem, we propose the resampling as well as reweighting schemes to iteratively adapt the model to the large intra-class variation of major class and balance the contributions to the loss from each class. Finally, we integrate these components together and devise a novel automatic inspection framework. Extensive experiments demonstrate that our proposed approaches are effective and can boost the performance compared to the baseline methods.
To accommodate rapid changes in the real world, the cognition system of humans is capable of cont... more To accommodate rapid changes in the real world, the cognition system of humans is capable of continually learning concepts. On the contrary, conventional deep learning models lack this capability of preserving previously learned knowledge. When a neural network is fine-tuned to learn new tasks, its performance on previously trained tasks will significantly deteriorate. Many recent works on incremental object detection tackle this problem by introducing advanced regularization. Although these methods have shown promising results, the benefits are often short-lived after the first incremental step. Under multi-step incremental learning, the trade-off between old knowledge preserving and new task learning becomes progressively more severe. Thus, the performance of regularization-based incremental object detectors gradually decays for subsequent learning steps. In this paper, we aim to alleviate this performance decay on multi-step incremental detection tasks by proposing a dilatable incremental object detector (DIODE). For the task-shared parameters, our method adaptively penalizes the changes of important weights for previous tasks. At the same time, the structure of the model is dilated or expanded by a limited number of task-specific parameters to promote new task learning. Extensive experiments on PASCAL VOC and COCO datasets demonstrate substantial improvements over the state-of-the-art methods. Notably, compared with the state-of-the-art methods, our method achieves up to 6.0% performance improvement by increasing the number of parameters by just 1.2% for each newly learned task.
Journal of Geophysical Research: Space Physics, 2021
By utilizing multipoint observations, we investigate the January 21, 2005 moderate storm's pr... more By utilizing multipoint observations, we investigate the January 21, 2005 moderate storm's prompt penetration electric field (PPEF) effects in the American sector, Rayleigh‐Taylor (R‐T) instability driven polarization electric (E) field effects in the African sector, and disturbance dynamo electric field (DDEF) effects in the Northern Hemisphere. We study how these E field effects impacted the equatorial ionization anomaly (EIA) and underlying forward fountain, the midlatitude trough, and the subauroral polarization streams (SAPS). In the daytime sector during the initial phase, the PPEF development was produced by solar wind pressure enhancements. In the dusk sector, the undershielding PEFs triggered R‐T instability mechanisms, which impacted the EIA during a prereversal enhancement (PRE) like scenario and afterwards. These led to the development of supersonic plasma bubbles drifting upward and westward. Supersonic bubbles are characteristics of superstorms, and their detection...
2014 1st Workshop on Pattern Recognition Techniques for Indirect Immunofluorescence Images, 2014
Manual microscopy analysis on specimens prepared by using Indirect Immunofluorescence protocol (I... more Manual microscopy analysis on specimens prepared by using Indirect Immunofluorescence protocol (IIF) is time consuming and labour intensive. This is due to the fact that the microscopic examination process involves human interpretation which tends to be subjective and prone to error. To that end, there has been a progressive and growing effort in the computer vision and pattern recognition scientific community to develop automated image-based analysis systems. Such systems process digitally-captured images from a patient specimen and produce a set of interpretations which may be incorporated into the human subjective analysis.
Journal of Geophysical Research: Space Physics, 2021
We investigate Kelvin‐Helmholtz (K‐H) waves on/near the magnetopause and surface waves near the p... more We investigate Kelvin‐Helmholtz (K‐H) waves on/near the magnetopause and surface waves near the plasmapause—in the outer region of the plasmasphere: in the hot zone—by utilizing multi‐instrument/satellite observations. Our main aim is to study how the K‐H waves and the K‐H instability mechanisms impacted the subauroral and auroral regions during the geomagnetic storms of May 27–29 and July 16, 2017. For the subauroral region, we specify the structured flows as Sub‐Auroral Polarization Streams Wave Structures (SAPS‐WS) and the combined flows—created by Abnormal Sub‐Auroral Ion Drifts (ASAID) and SAID or SAPS—as a complex equatorward‐poleward ASAID‐SAID or SAPS‐ASAID. For the auroral zone, we identify the large auroral undulations appearing inside the auroral zone. The correlated observations of the K‐H waves, the structured or complex subauroral flows and large auroral undulations, and the local geomagnetic field oscillations confirm the connections of both the subauroral flows and t...
The human vision and perception system is inherently incremental where new knowledge is continual... more The human vision and perception system is inherently incremental where new knowledge is continually learned over time whilst existing knowledge is retained. On the other hand, deep learning networks are ill-equipped for incremental learning. When a well-trained network is adapted to new categories, its performance on the old categories will dramatically degrade. To address this problem, incremental learning methods have been explored which preserve the old knowledge of deep learning models. However, the state-of-the-art incremental object detector employs an external fixed region proposal method that increases overall computation time and reduces accuracy comparing to Region Proposal Network (RPN) based object detectors such as Faster RCNN. The purpose of this paper is to design an efficient end-to-end incremental object detector using knowledge distillation. We first evaluate and analyze the performance of the RPN-based detector with classic distillation on incremental detection tasks. Then, we introduce multi-network adaptive distillation that properly retains knowledge from the old categories when fine-tuning the model for new task. Experiments on the benchmark datasets, PASCAL VOC and COCO, demonstrate that the proposed incremental detector based on Faster RCNN is more accurate as well as being 13 times faster than the baseline detector.
Despite much progress being made in the field of object recognition with the advances of deep lea... more Despite much progress being made in the field of object recognition with the advances of deep learning, there are still several factors negatively affecting the performance of deep learning models. Domain shift is one of these factors and is caused by discrepancies in the distributions of the testing and training data. In this paper, we focus on the problem of compact feature clustering in domain generalization to help optimize the embedding space from multi-domain data. We design a domainaware triplet loss for domain generalization to help the model to not only cluster similar semantic features, but also to disperse features arising from the domain. Unlike previous methods focusing on distribution alignment, our algorithm is designed to disperse domain information in the embedding space. The basic idea is motivated based on the assumption that embedding features can be clustered based on domain information, which is mathematically and empirically supported in this paper. In addition, during our exploration of feature clustering in domain generalization, we note that factors affecting the convergence of metric learning loss in domain generalization are more important than the pre-defined domains. To solve this issue, we utilize two methods to normalize the embedding space, reducing the internal covariate shift of the embedding features. The ablation study demonstrates the effectiveness of our algorithm. Moreover, the experiments on the benchmark datasets, including PACS, VLCS and Office-Home, show that our method outperforms related methods focusing on domain discrepancy. In particular, our results on RegnetY-16 are significantly better than state-of-the-art methods on the benchmark datasets. Our code will be released at https://github.com/workerbcd/DCT.
We study the newly formed (with an upper limit of ∼101 min) subauroral flows in the topside ionos... more We study the newly formed (with an upper limit of ∼101 min) subauroral flows in the topside ionosphere. These include the dawnside subauroral flows observed at ∼5–6 magnetic local time (MLT), the duskside sub‐auroral polarization streams (SAPS) at ∼15 MLT, and the nightside sub‐auroral ion drifts (SAID) at ∼19–20 MLT. We employ the same types of observations for investigating the various underlying magnetospheric processes related to the development of the newly formed subauroral flows. These processes include (a) the stretching of magnetic field lines (leading to plasma density dropouts) due to the unfolding magnetotail reconnection, (b) the dawnside and nightside diversions of the magnetotail‐reconnection—related earthward plasma flows, (c) the ongoing particle injections, and (d) the trapping of ring current protons at subauroral and auroral latitudes. Based on these processes (a–d), we found strong similarities in the development of the newly formed dawnside and duskside/nightsi...
The continual appearance of new objects in the visual world poses considerable challenges for cur... more The continual appearance of new objects in the visual world poses considerable challenges for current deep learning methods in realworld deployments. The challenge of new task learning is often exacerbated by the scarcity of data for the new categories due to rarity or cost. Here we explore the important task of Few-Shot Class-Incremental Learning (FSCIL) and its extreme data scarcity condition of one-shot. An ideal FSCIL model needs to perform well on all classes, regardless of their presentation order or paucity of data. It also needs to be robust to open-set real-world conditions and be easily adapted to the new tasks that always arise in the field. In this paper, we first reevaluate the current task setting and propose a more comprehensive and practical setting for the FSCIL task. Then, inspired by the similarity of the goals for FSCIL and modern face recognition systems, we propose our method-Augmented Angular Loss Incremental Classification or ALICE. In ALICE, instead of the commonly used cross-entropy loss, we propose to use the angular penalty loss to obtain well-clustered features. As the obtained features not only need to be compactly clustered but also diverse enough to maintain generalization for future incremental classes, we further discuss how class augmentation, data augmentation, and data balancing affect classification performance. Experiments on benchmark datasets, including CIFAR100, miniImageNet, and CUB200, demonstrate the improved performance of ALICE over the state-of-the-art FSCIL methods.
2019 Digital Image Computing: Techniques and Applications (DICTA)
Galvanised steel transmission towers in electrical power grids suffer from corrosion to different... more Galvanised steel transmission towers in electrical power grids suffer from corrosion to different levels depending on age and environment. To ensure the power grid can operate safely, significant resources are spent to monitor the corrosion level of towers. Photographs from helicopters, drones, and a variety of staffs are often used to capture condition, however, these images still need manual inspection to determine the corrosion level before carrying out maintenance works. In this paper, we describe a framework employing multiple deep neural networks based classifiers and detectors to perform automatic image-based condition monitoring for steel transmission towers. Given a random variety of images of a tower, our proposed framework will first determine the location of the image on the structure via a trained zone classifier. Then, fine-grain corrosion inspection will be performed on both fasteners and structural members, respectively. In addition, an automatic zoomin functionality will be applied to images which have high resolution but are a long distance away. This step will ensure the detection performance on small objects on the tower. Finally, the overall corrosion status report for this tower will be calculated and generated automatically. Additionally, we released a subset of our data to contribute to this novel direction. Experiments show that our framework can assess the tower efficiently.
2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2017
Gram stained direct smears test is clinically useful in early identification of infections. Unfor... more Gram stained direct smears test is clinically useful in early identification of infections. Unfortunately, this practice is considered time consuming and labour intensive. Most existing effort in this area is to perform highmagnification analysis of images taken from manually selected areas. In this paper, we address the problem of the automatic selection of candidate areas (or patches) for subsequent high-magnification analysis. Specifically, we explore and study the possibility of selecting good working areas based on low-magnification images, where bacteria are likely to be found when viewed in high-magnification images. To this end, we develop an approach to classify the areas of interest according to the textural information of the image patch. We explore and study the efficacy of traditional textural features such as Histogram of Gradients, Local Binary Patterns, and 2 Dimensional Discrete Cosine Transform. Experiments show that the best variant method is able to select working areas where it is likely to find bacteria in high-powered objective images in a wide range of images.
Pathology tests play a major role in our health care system and they assist physicians to make ac... more Pathology tests play a major role in our health care system and they assist physicians to make accurate diagnoses. Recently, there has been growing interest in developing Computer-Aided Diagnostic (CAD) systems which should further improve the reliability and consistency of these tests. We discuss the Anti-Nuclear Antibodies (ANA) test via Indirect Immuno uorescence protocol on Human Epithelial Type 2 (HEp- 2) cells which is normally used to determine the presence of Connective Tissue Diseases such as Sjffogren's syndrome, Systemic Lupus Erythematosus and Rheumatoid Arthritis. One task in the test is to determine the pattern of positive HEp-2 cells. This chapter discusses and compares both traditional and contemporary approaches to this problem via image-based analysis. Whilst, the traditional approaches comprise several popular methods used to address generic image classification problems, the contemporary approaches apply advanced mathematical techniques to improve system accu...
2016 23rd International Conference on Pattern Recognition (ICPR), 2016
Automatically recognising facial emotions has drawn increasing attention in computer vision. Faci... more Automatically recognising facial emotions has drawn increasing attention in computer vision. Facial landmark based methods are one of the most widely used approaches to perform this task. However, these approaches do not provide good performance. Thus, researchers usually tend to combine more information such as textural and audio information to increase the recognition rate. In this paper we propose a novel method, here called the landmark manifold, that shows the possibility to achieve competitive performance by facial landmark information alone. Through experiments on the well-known dataset: marked Cohn-Kanade extended facial emotion dataset (CK+), we show that with accurate facial landmarks, our simple approach is fast to run and can achieve competitive performance with enormously expensive methods.
2019 Digital Image Computing: Techniques and Applications (DICTA), 2019
The condition of the Glomeruli, or filter sacks, in renal Direct Immunofluorescence (DIF) specime... more The condition of the Glomeruli, or filter sacks, in renal Direct Immunofluorescence (DIF) specimens is a critical indicator for diagnosing kidney diseases. A digital pathology system which digitizes a glass histology slide into a Whole Slide Image (WSI) and then automatically detects and zooms in on the glomeruli with a higher magnification objective will be extremely helpful for pathologists. In this paper, using glomerulus detection as the study case, we provide analysis and observations on several important issues to help with the development of Computer Aided Diagnostic (CAD) systems to process WSIs. Large image resolution, large file size, and data scarcity are always challenging to deal with. To this end, we first examine image downsampling rates in terms of their effect on detection accuracy. Second, we examine the impact of image compression. Third, we examine the relationship between the size of the training set and detection accuracy. To understand the above issues, experiments are performed on the state-of-the-art detectors: Faster R-CNN, R-FCN, Mask R-CNN and SSD. Critical findings are observed: (1) The best balance between detection accuracy, detection speed and file size is achieved at 8 times downsampling captured with a 40× objective; (2) compression which reduces the file size dramatically, does not necessarily have an adverse effect on overall accuracy; (3) reducing the amount of training data to some extents causes a drop in precision but has a negligible impact on the recall; (4) in most cases, Faster R-CNN achieves the best accuracy in the glomerulus detection task. We show that the image file size of 40× WSI images can be reduced by a factor of over 6000 with negligible loss of glomerulus detection accuracy.
2019 IEEE International Conference on Image Processing (ICIP), 2019
missing texture and edge information. In this paper, we proposed a framework to translate images ... more missing texture and edge information. In this paper, we proposed a framework to translate images while preserving more realistic textures and details. To this end, we disentangle the samples into shared content space and domain-specific style domain. Then, according to the blurred outlines and textures in the source domain, we introduce the classic canny edge detection algorithm to encode the boundary and edge information in the content latent space. We test the proposed method in the thermal to visible image translation scenario and the experimental results demonstrate that the proposed method outperforms several other state-of-the-art models.
2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020
The difficulty of processing gigapixel whole slide images (WSIs) in clinical microscopy has been ... more The difficulty of processing gigapixel whole slide images (WSIs) in clinical microscopy has been a long-standing barrier to implementing computer aided diagnostic systems. Since modern computing resources are unable to perform computations at this extremely large scale, current state of the art methods utilize patch-based processing to preserve the resolution of WSIs. However, these methods are often resource intensive and make significant compromises on processing time. In this paper, we demonstrate that conventional patch-based processing is redundant for certain WSI classification tasks where high resolution is only required in a minority of cases. This reflects what is observed in clinical practice; where a pathologist may screen slides using a low power objective and only switch to a high power in cases where they are uncertain about their findings. To eliminate these redundancies, we propose a method for the selective use of high resolution processing based on the confidence of predictions on downscaled WSIs-we call this the Selective Objective Switch (SOS). Our method is validated on a novel dataset of 684 Liver-Kidney-Stomach immunofluorescence WSIs routinely used in the investigation of autoimmune liver disease. By limiting high resolution processing to cases which cannot be classified confidently at low resolution, we maintain the accuracy of patch-level analysis whilst reducing the inference time by a factor of 7.74.
A method to segment a nonstationary signal into non-overlapping successive near stationary segmen... more A method to segment a nonstationary signal into non-overlapping successive near stationary segments using the Covariance method of Autoregressive (AR) Modelling is presented. An assumption of" local ergodicityn within the near stationary segments allows improved spectral estimation because of the adaptive nature of the windowing_ Performance is evaluated on simulated AR data and real oceanographic signals.
2019 IEEE International Conference on Image Processing (ICIP)
Electrical distribution poles are important assets in electricity supply. These poles need to be ... more Electrical distribution poles are important assets in electricity supply. These poles need to be maintained in good condition to ensure they protect community safety, maintain reliability of supply, and meet legislative obligations. However, maintaining such a large volumes of assets is an expensive and challenging task. To address this, recent approaches utilise imagery data captured from helicopter and/or drone inspections. Whilst reducing the cost for manual inspection, manual analysis on each image is still required. As such, several image-based automated inspection systems have been proposed. In this paper, we target two major challenges: tiny object detection and extremely imbalanced datasets, which currently hinder the wide deployment of the automatic inspection. We propose a novel two-stage zoom-in detection method to gradually focus on the object of interest. To address the imbalanced dataset problem, we propose the resampling as well as reweighting schemes to iteratively adapt the model to the large intra-class variation of major class and balance the contributions to the loss from each class. Finally, we integrate these components together and devise a novel automatic inspection framework. Extensive experiments demonstrate that our proposed approaches are effective and can boost the performance compared to the baseline methods.
To accommodate rapid changes in the real world, the cognition system of humans is capable of cont... more To accommodate rapid changes in the real world, the cognition system of humans is capable of continually learning concepts. On the contrary, conventional deep learning models lack this capability of preserving previously learned knowledge. When a neural network is fine-tuned to learn new tasks, its performance on previously trained tasks will significantly deteriorate. Many recent works on incremental object detection tackle this problem by introducing advanced regularization. Although these methods have shown promising results, the benefits are often short-lived after the first incremental step. Under multi-step incremental learning, the trade-off between old knowledge preserving and new task learning becomes progressively more severe. Thus, the performance of regularization-based incremental object detectors gradually decays for subsequent learning steps. In this paper, we aim to alleviate this performance decay on multi-step incremental detection tasks by proposing a dilatable incremental object detector (DIODE). For the task-shared parameters, our method adaptively penalizes the changes of important weights for previous tasks. At the same time, the structure of the model is dilated or expanded by a limited number of task-specific parameters to promote new task learning. Extensive experiments on PASCAL VOC and COCO datasets demonstrate substantial improvements over the state-of-the-art methods. Notably, compared with the state-of-the-art methods, our method achieves up to 6.0% performance improvement by increasing the number of parameters by just 1.2% for each newly learned task.
Journal of Geophysical Research: Space Physics, 2021
By utilizing multipoint observations, we investigate the January 21, 2005 moderate storm's pr... more By utilizing multipoint observations, we investigate the January 21, 2005 moderate storm's prompt penetration electric field (PPEF) effects in the American sector, Rayleigh‐Taylor (R‐T) instability driven polarization electric (E) field effects in the African sector, and disturbance dynamo electric field (DDEF) effects in the Northern Hemisphere. We study how these E field effects impacted the equatorial ionization anomaly (EIA) and underlying forward fountain, the midlatitude trough, and the subauroral polarization streams (SAPS). In the daytime sector during the initial phase, the PPEF development was produced by solar wind pressure enhancements. In the dusk sector, the undershielding PEFs triggered R‐T instability mechanisms, which impacted the EIA during a prereversal enhancement (PRE) like scenario and afterwards. These led to the development of supersonic plasma bubbles drifting upward and westward. Supersonic bubbles are characteristics of superstorms, and their detection...
2014 1st Workshop on Pattern Recognition Techniques for Indirect Immunofluorescence Images, 2014
Manual microscopy analysis on specimens prepared by using Indirect Immunofluorescence protocol (I... more Manual microscopy analysis on specimens prepared by using Indirect Immunofluorescence protocol (IIF) is time consuming and labour intensive. This is due to the fact that the microscopic examination process involves human interpretation which tends to be subjective and prone to error. To that end, there has been a progressive and growing effort in the computer vision and pattern recognition scientific community to develop automated image-based analysis systems. Such systems process digitally-captured images from a patient specimen and produce a set of interpretations which may be incorporated into the human subjective analysis.
Journal of Geophysical Research: Space Physics, 2021
We investigate Kelvin‐Helmholtz (K‐H) waves on/near the magnetopause and surface waves near the p... more We investigate Kelvin‐Helmholtz (K‐H) waves on/near the magnetopause and surface waves near the plasmapause—in the outer region of the plasmasphere: in the hot zone—by utilizing multi‐instrument/satellite observations. Our main aim is to study how the K‐H waves and the K‐H instability mechanisms impacted the subauroral and auroral regions during the geomagnetic storms of May 27–29 and July 16, 2017. For the subauroral region, we specify the structured flows as Sub‐Auroral Polarization Streams Wave Structures (SAPS‐WS) and the combined flows—created by Abnormal Sub‐Auroral Ion Drifts (ASAID) and SAID or SAPS—as a complex equatorward‐poleward ASAID‐SAID or SAPS‐ASAID. For the auroral zone, we identify the large auroral undulations appearing inside the auroral zone. The correlated observations of the K‐H waves, the structured or complex subauroral flows and large auroral undulations, and the local geomagnetic field oscillations confirm the connections of both the subauroral flows and t...
Uploads
Papers by Brian Lovell