Although deep neural networks (NNs) have achievedstate-of-the-art accuracy in many visual recognition tasks,the growing computational complexity and energy con-sumption of networks remains an issue, especially for ap-plications on platforms with limited resources and requir-ing real-time processing. Filter pruning techniques haverecently shown promising results for the compression andacceleration of convolutional NNs (CNNs). However, thesetechniques involve numerous steps and complex optimisa-tions because some only prune after training CNNs, whileothers prune from scratch during training by integratingsparsity constraints or modifying the loss function.In this paper we propose a new Progressive GradientPruning (PGP) technique for iterative filter pruning dur-ing training. In contrast to previous progressive pruningtechniques, it relies on a novel filter selection criterion thatmeasures the change in filter weights, uses a new hard andsoft pruning strategy and effectively adapts momentum ten-sors during the backward propagation pass. Experimentalresults obtained after training various CNNs on image datafor classification, object detection and domain adaptationbenchmarks indicate that the PGP technique can achievea better trade-off between classification accuracy and net-work (time and memory) complexity than PSFP and otherstate-of-the-art filter pruning techniques.
In this paper, we propose a novel classification scheme for the remotely sensed hyperspectral image (HSI), namely SP-DLRR, by comprehensively exploring its unique characteristics, including the local spatial information and low-rankness. SP-DLRR is mainly composed of two modules, i.e., the classification-guided superpixel segmentation and the discriminative low-rank representation, which are iteratively conducted. Specifically, by utilizing the local spatial information and incorporating the predictions from a typical classifier, the first module segments pixels of an input HSI (or its restoration generated by the second module) into superpixels. According to the resulting superpixels, the pixels of the input HSI are then grouped into clusters and fed into our novel discriminative low-rank representation model with an effective numerical solution. Such a model is capable of increasing the intra-class similarity by suppressing the spectral variations locally while promoting the inter-class discriminability globally, leading to a restored HSI with more discriminative pixels. Experimental results on three benchmark datasets demonstrate the significant superiority of SP-DLRR over state-of-the-art methods, especially for the case with an extremely limited number of training pixels.
The automated Interstitial Lung Diseases (ILDs) classification technique is essential for assisting clinicians during the diagnosis process. Detecting and classifying ILDs patterns is a challenging problem. This paper introduces an end-to-end deep convolution neural network (CNN) for classifying ILDs patterns. The proposed model comprises four convolutional layers with different kernel sizes and Rectified Linear Unit (ReLU) activation function, followed by batch normalization and max-pooling with a size equal to the final feature map size well as four dense layers. We used the ADAM optimizer to minimize categorical cross-entropy. A dataset consisting of 21328 image patches of 128 CT scans with five classes is taken to train and assess the proposed model. A comparison study showed that the presented model outperformed pre-trained CNNs and five-fold cross-validation on the same dataset. For ILDs pattern classification, the proposed approach achieved the accuracy scores of 99.09% and the average F score of 97.9%, outperforming three pre-trained CNNs. These outcomes show that the proposed model is relatively state-of-the-art in precision, recall, f score, and accuracy.
Electrocardiogram (ECG) is a simple non-invasive measure to identify heart-related issues such as irregular heartbeats known as arrhythmias. While artificial intelligence and machine learning is being utilized in a wide range of healthcare related applications and datasets, many arrhythmia classifiers using deep learning methods have been proposed in recent years. However, sizes of the available datasets from which to build and assess machine learning models is often very small and the lack of well-annotated public ECG datasets is evident. In this paper, we propose a deep transfer learning framework that is aimed to perform classification on a small size training dataset. The proposed method is to fine-tune a general-purpose image classifier ResNet-18 with MIT-BIH arrhythmia dataset in accordance with the AAMI EC57 standard. This paper further investigates many existing deep learning models that have failed to avoid data leakage against AAMI recommendations. We compare how different data split methods impact the model performance. This comparison study implies that future work in arrhythmia classification should follow the AAMI EC57 standard when using any including MIT-BIH arrhythmia dataset.
Text role classification involves classifying the semantic role of textual elements within scientific charts. For this task, we propose to finetune two pretrained multimodal document layout analysis models, LayoutLMv3 and UDOP, on chart datasets. The transformers utilize the three modalities of text, image, and layout as input. We further investigate whether data augmentation and balancing methods help the performance of the models. The models are evaluated on various chart datasets, and results show that LayoutLMv3 outperforms UDOP in all experiments. LayoutLMv3 achieves the highest F1-macro score of 82.87 on the ICPR22 test dataset, beating the best-performing model from the ICPR22 CHART-Infographics challenge. Moreover, the robustness of the models is tested on a synthetic noisy dataset ICPR22-N. Finally, the generalizability of the models is evaluated on three chart datasets, CHIME-R, DeGruyter, and EconBiz, for which we added labels for the text roles. Findings indicate that even in cases where there is limited training data, transformers can be used with the help of data augmentation and balancing methods. The source code and datasets are available on GitHub under https://github.com/hjkimk/text-role-classification
Brain tumor is a life-threatening problem and hampers the normal functioning of the human body. The average five-year relative survival rate for malignant brain tumors is 35.6 percent. For proper diagnosis and efficient treatment planning, it is necessary to detect the brain tumor in early stages. Due to advancement in medical imaging technology, the brain images are taken in different modalities. The ability to extract relevant characteristics from magnetic resonance imaging (MRI) scans is a crucial step for brain tumor classifiers. Several studies have proposed various strategies to extract relevant features from different modalities of MRI to predict the growth of abnormal tumors. Most techniques used conventional methods of image processing for feature extraction and machine learning for classification. More recently, the use of deep learning algorithms in medical imaging has resulted in significant improvements in the classification and diagnosis of brain tumors. Since tumors are located at different regions of the brain, localizing the tumor and classifying it to a particular category is a challenging task. The objective of this project is to develop a predictive system for brain tumor detection using machine learning(ensembling).
Recent advances in prototypical learning have shown remarkable potential to provide useful decision interpretations associating activation maps and predictions with class-specific training prototypes. Such prototypical learning has been well-studied for various single-label diseases, but for quite relevant and more challenging multi-label diagnosis, where multiple diseases are often concurrent within an image, existing prototypical learning models struggle to obtain meaningful activation maps and effective class prototypes due to the entanglement of the multiple diseases. In this paper, we present a novel Cross- and Intra-image Prototypical Learning (CIPL) framework, for accurate multi-label disease diagnosis and interpretation from medical images. CIPL takes advantage of common cross-image semantics to disentangle the multiple diseases when learning the prototypes, allowing a comprehensive understanding of complicated pathological lesions. Furthermore, we propose a new two-level alignment-based regularisation strategy that effectively leverages consistent intra-image information to enhance interpretation robustness and predictive performance. Extensive experiments show that our CIPL attains the state-of-the-art (SOTA) classification accuracy in two public multi-label benchmarks of disease diagnosis: thoracic radiography and fundus images. Quantitative interpretability results show that CIPL also has superiority in weakly-supervised thoracic disease localisation over other leading saliency- and prototype-based explanation methods.
Vehicle classification is a hot computer vision topic, with studies ranging from ground-view up to top-view imagery. In remote sensing, the usage of top-view images allows for understanding city patterns, vehicle concentration, traffic management, and others. However, there are some difficulties when aiming for pixel-wise classification: (a) most vehicle classification studies use object detection methods, and most publicly available datasets are designed for this task, (b) creating instance segmentation datasets is laborious, and (c) traditional instance segmentation methods underperform on this task since the objects are small. Thus, the present research objectives are: (1) propose a novel semi-supervised iterative learning approach using GIS software, (2) propose a box-free instance segmentation approach, and (3) provide a city-scale vehicle dataset. The iterative learning procedure considered: (1) label a small number of vehicles, (2) train on those samples, (3) use the model to classify the entire image, (4) convert the image prediction into a polygon shapefile, (5) correct some areas with errors and include them in the training data, and (6) repeat until results are satisfactory. To separate instances, we considered vehicle interior and vehicle borders, and the DL model was the U-net with the Efficient-net-B7 backbone. When removing the borders, the vehicle interior becomes isolated, allowing for unique object identification. To recover the deleted 1-pixel borders, we proposed a simple method to expand each prediction. The results show better pixel-wise metrics when compared to the Mask-RCNN (82% against 67% in IoU). On per-object analysis, the overall accuracy, precision, and recall were greater than 90%. This pipeline applies to any remote sensing target, being very efficient for segmentation and generating datasets.
The partially occluded image recognition (POIR) problem has been a challenge for artificial intelligence for a long time. A common strategy to handle the POIR problem is using the non-occluded features for classification. Unfortunately, this strategy will lose effectiveness when the image is severely occluded, since the visible parts can only provide limited information. Several studies in neuroscience reveal that feature restoration which fills in the occluded information and is called amodal completion is essential for human brains to recognize partially occluded images. However, feature restoration is commonly ignored by CNNs, which may be the reason why CNNs are ineffective for the POIR problem. Inspired by this, we propose a novel brain-inspired feature restoration network (BIFRNet) to solve the POIR problem. It mimics a ventral visual pathway to extract image features and a dorsal visual pathway to distinguish occluded and visible image regions. In addition, it also uses a knowledge module to store object prior knowledge and uses a completion module to restore occluded features based on visible features and prior knowledge. Thorough experiments on synthetic and real-world occluded image datasets show that BIFRNet outperforms the existing methods in solving the POIR problem. Especially for severely occluded images, BIRFRNet surpasses other methods by a large margin and is close to the human brain performance. Furthermore, the brain-inspired design makes BIFRNet more interpretable.
In the realm of image quantization exemplified by VQGAN, the process encodes images into discrete tokens drawn from a codebook with a predefined size. Recent advancements, particularly with LLAMA 3, reveal that enlarging the codebook significantly enhances model performance. However, VQGAN and its derivatives, such as VQGAN-FC (Factorized Codes) and VQGAN-EMA, continue to grapple with challenges related to expanding the codebook size and enhancing codebook utilization. For instance, VQGAN-FC is restricted to learning a codebook with a maximum size of 16,384, maintaining a typically low utilization rate of less than 12% on ImageNet. In this work, we propose a novel image quantization model named VQGAN-LC (Large Codebook), which extends the codebook size to 100,000, achieving an utilization rate exceeding 99%. Unlike previous methods that optimize each codebook entry, our approach begins with a codebook initialized with 100,000 features extracted by a pre-trained vision encoder. Optimization then focuses on training a projector that aligns the entire codebook with the feature distributions of the encoder in VQGAN-LC. We demonstrate the superior performance of our model over its counterparts across a variety of tasks, including image reconstruction, image classification, auto-regressive image generation using GPT, and image creation with diffusion- and flow-based generative models. Code and models are available at https://github.com/zh460045050/VQGAN-LC.
Zero-shot sketch-based image retrieval (SBIR) is an emerging task in computer vision, allowing to retrieve natural images relevant to sketch queries that might not been seen in the training phase. Existing works either require aligned sketch-image pairs or inefficient memory fusion layer for mapping the visual information to a semantic space. In this work, we propose a semantically aligned paired cycle-consistent generative (SEM-PCYC) model for zero-shot SBIR, where each branch maps the visual information to a common semantic space via an adversarial training. Each of these branches maintains a cycle consistency that only requires supervision at category levels, and avoids the need of highly-priced aligned sketch-image pairs. A classification criteria on the generators' outputs ensures the visual to semantic space mapping to be discriminating. Furthermore, we propose to combine textual and hierarchical side information via a feature selection auto-encoder that selects discriminating side information within a same end-to-end model. Our results demonstrate a significant boost in zero-shot SBIR performance over the state-of-the-art on the challenging Sketchy and TU-Berlin datasets.
With the thriving of deep learning, 3D Convolutional Neural Networks have become a popular choice in volumetric image analysis due to their impressive 3D contexts mining ability. However, the 3D convolutional kernels will introduce a significant increase in the amount of trainable parameters. Considering the training data is often limited in biomedical tasks, a tradeoff has to be made between model size and its representational power. To address this concern, in this paper, we propose a novel 3D Dense Separated Convolution (3D-DSC) module to replace the original 3D convolutional kernels. The 3D-DSC module is constructed by a series of densely connected 1D filters. The decomposition of 3D kernel into 1D filters reduces the risk of over-fitting by removing the redundancy of 3D kernels in a topologically constrained manner, while providing the infrastructure for deepening the network. By further introducing nonlinear layers and dense connections between 1D filters, the network's representational power can be significantly improved while maintaining a compact architecture. We demonstrate the superiority of 3D-DSC on volumetric image classification and segmentation, which are two challenging tasks often encountered in biomedical image computing.
In adversarial attacks intended to confound deep learning models, most studies have focused on limiting the magnitude of the modification so that humans do not notice the attack. On the other hand, during an attack against autonomous cars, for example, most drivers would not find it strange if a small insect image were placed on a stop sign, or they may overlook it. In this paper, we present a systematic approach to generate natural adversarial examples against classification models by employing such natural-appearing perturbations that imitate a certain object or signal. We first show the feasibility of this approach in an attack against an image classifier by employing generative adversarial networks that produce image patches that have the appearance of a natural object to fool the target model. We also introduce an algorithm to optimize placement of the perturbation in accordance with the input image, which makes the generation of adversarial examples fast and likely to succeed. Moreover, we experimentally show that the proposed approach can be extended to the audio domain, for example, to generate perturbations that sound like the chirping of birds to fool a speech classifier.
In this paper we present a novel radar-camera sensor fusion framework for accurate object detection and distance estimation in autonomous driving scenarios. The proposed architecture uses a middle-fusion approach to fuse the radar point clouds and RGB images. Our radar object proposal network uses radar point clouds to generate 3D proposals from a set of 3D prior boxes. These proposals are mapped to the image and fed into a Radar Proposal Refinement (RPR) network for objectness score prediction and box refinement. The RPR network utilizes both radar information and image feature maps to generate accurate object proposals and distance estimations. The radar-based proposals are combined with image-based proposals generated by a modified Region Proposal Network (RPN). The RPN has a distance regression layer for estimating distance for every generated proposal. The radar-based and image-based proposals are merged and used in the next stage for object classification. Experiments on the challenging nuScenes dataset show our method outperforms other existing radar-camera fusion methods in the 2D object detection task while at the same time accurately estimates objects' distances.
Multimodal machine learning, especially text-to-image models like Stable Diffusion and DALL-E 3, has gained significance for transforming text into detailed images. Despite their growing use and remarkable generative capabilities, there is a pressing need for a detailed examination of these models' behavior, particularly with respect to memorization. Historically, memorization in machine learning has been context-dependent, with diverse definitions emerging from classification tasks to complex models like Large Language Models (LLMs) and Diffusion models. Yet, a definitive concept of memorization that aligns with the intricacies of text-to-image synthesis remains elusive. This understanding is vital as memorization poses privacy risks yet is essential for meeting user expectations, especially when generating representations of underrepresented entities. In this paper, we introduce a specialized definition of memorization tailored to text-to-image models, categorizing it into three distinct types according to user expectations. We closely examine the subtle distinctions between intended and unintended memorization, emphasizing the importance of balancing user privacy with the generative quality of the model outputs. Using the Stable Diffusion model, we offer examples to validate our memorization definitions and clarify their application.
A modified version of MRFFCM (Markov Random Field Fuzzy C means) based SAR (Synthetic aperture Radar) image change detection method is proposed in this paper. It involves three steps: Difference Image (DI) generation by using Gauss-log ratio operator, speckle noise reduction by SRAD (Speckle Reducing Anisotropic Diffusion), and the detection of changed regions by using MRFFCM. The proposed method is compared with existing methods such as FCM and MRFFCM using simulated and real SAR images. The measures used for evaluation includes Overall Error (OE), Percentage Correct Classification (PCC), Kappa Coefficient (KC), Root Mean Square Error (RMSE), and Peak Signal to Noise Ratio (PSNR). The results show that the proposed method is better compared to FCM and MRFFCM based change detection method.
Convolutional neural nets (convnets) trained from massive labeled datasets have substantially improved the state-of-the-art in image classification and object detection. However, visual understanding requires establishing correspondence on a finer level than object category. Given their large pooling regions and training from whole-image labels, it is not clear that convnets derive their success from an accurate correspondence model which could be used for precise localization. In this paper, we study the effectiveness of convnet activation features for tasks requiring correspondence. We present evidence that convnet features localize at a much finer scale than their receptive field sizes, that they can be used to perform intraclass alignment as well as conventional hand-engineered features, and that they outperform conventional features in keypoint prediction on objects from PASCAL VOC 2011.
Identification and classification of environmental forensics, with the petroleum forensics as the main application, requires an effective technology or method to distinguish between the closely located forensics as they share many main biomarkers. Two-dimensional gas chromatography is one of these technologies with which a petroleum forensic is separated into its chemical compounds, resulting in a three-dimensional image, GCXGC image. Therefore, distinguishing between two petroleum forensics is equivalent to the comparison between their corresponding GCXGC images. In this paper, we present a technique, called Quantized Peak Topography Map (QPTM), which results in a better separation between the GCXGC images. We validate our proposed method on a model dataset, consisting of thirtyfour GCXGC images, extracted from the different parts of the world.
Weakly-supervised classification of histopathology slides is a computationally intensive task, with a typical whole slide image (WSI) containing billions of pixels to process. We propose Discriminative Region Active Sampling for Multiple Instance Learning (DRAS-MIL), a computationally efficient slide classification method using attention scores to focus sampling on highly discriminative regions. We apply this to the diagnosis of ovarian cancer histological subtypes, which is an essential part of the patient care pathway as different subtypes have different genetic and molecular profiles, treatment options, and patient outcomes. We use a dataset of 714 WSIs acquired from 147 epithelial ovarian cancer patients at Leeds Teaching Hospitals NHS Trust to distinguish the most common subtype, high-grade serous carcinoma, from the other four subtypes (low-grade serous, endometrioid, clear cell, and mucinous carcinomas) combined. We demonstrate that DRAS-MIL can achieve similar classification performance to exhaustive slide analysis, with a 3-fold cross-validated AUC of 0.8679 compared to 0.8781 with standard attention-based MIL classification. Our approach uses at most 18% as much memory as the standard approach, while taking 33% of the time when evaluating on a GPU and only 14% on a CPU alone. Reducing prediction time and memory requirements may benefit clinical deployment and the democratisation of AI, reducing the extent to which computational hardware limits end-user adoption.
Automatic breast cancer classification in histopathology images is crucial for precise diagnosis and treatment planning. Recently, classification approaches based on the ResNet architecture have gained popularity for significantly improving accuracy by using skip connections to mitigate vanishing gradient problems, thereby integrating low-level and high-level feature information. Nevertheless, the conventional ResNet architecture faces challenges such as data imbalance and limited interpretability, necessitating cross-domain knowledge and collaboration among medical experts. This study effectively addresses these challenges by introducing a novel method for breast cancer classification, the Dual-Activated Lightweight Attention ResNet50 (DALAResNet50) model. It integrates a pre-trained ResNet50 model with a lightweight attention mechanism, embedding an attention module in the fourth layer of ResNet50 and incorporating two fully connected layers with LeakyReLU and ReLU activation functions to enhance feature learning capabilities. The DALAResNet50 method was tested on breast cancer histopathology images from the BreakHis Database across magnification factors of 40X, 100X, 200X, and 400X, achieving accuracies of 98.5%, 98.7%, 97.9%, and 94.3%, respectively. It was also compared with established deep learning models such as SEResNet50, DenseNet121, VGG16, VGG16Inception, ViT, Swin-Transformer, Dinov2_Vitb14, and ResNet50. The reported results of DALAResNet50 have been shown to outperform the compared approaches regarding accuracy, F1 score, IBA, and GMean, demonstrating significant robustness and broad applicability when dealing with different magnifications and imbalanced breast cancer datasets