Deploying high-performance vision transformer (ViT) models on ubiquitous Internet of Things (IoT) devices to provide high-quality vision services will revolutionize the way we live, work, and interact with the world. Due to the contradiction between the limited resources of IoT devices and resource-intensive ViT models, the use of cloud servers to assist ViT model training has become mainstream. However, due to the larger number of parameters and floating-point operations (FLOPs) of the existing ViT models, the model parameters transmitted by cloud servers are large and difficult to run on resource-constrained IoT devices. To this end, this paper proposes a transmission-friendly ViT model, TFormer, for deployment on resource-constrained IoT devices with the assistance of a cloud server. The high performance and small number of model parameters and FLOPs of TFormer are attributed to the proposed hybrid layer and the proposed partially connected feed-forward network (PCS-FFN). The hybrid layer consists of nonlearnable modules and a pointwise convolution, which can obtain multitype and multiscale features with only a few parameters and FLOPs to improve the TFormer performance. The PCS-FFN adopts group convolution to reduce the number of parameters. The key idea of this paper is to propose TFormer with few model parameters and FLOPs to facilitate applications running on resource-constrained IoT devices to benefit from the high performance of the ViT models. Experimental results on the ImageNet-1K, MS COCO, and ADE20K datasets for image classification, object detection, and semantic segmentation tasks demonstrate that the proposed model outperforms other state-of-the-art models. Specifically, TFormer-S achieves 5% higher accuracy on ImageNet-1K than ResNet18 with 1.4$\times$ fewer parameters and FLOPs.
Annually 8500 neonatal deaths are reported in the US due to respiratory failure. Recently, Lung Ultrasound (LUS), due to its radiation free nature, portability, and being cheaper is gaining wide acceptability as a diagnostic tool for lung conditions. However, lack of highly trained medical professionals has limited its use especially in remote areas. To address this, an automated screening system that captures characteristics of the LUS patterns can be of significant assistance to clinicians who are not experts in lung ultrasound (LUS) images. In this paper, we propose a feature extraction method designed to quantify the spatially-localized line patterns and texture patterns found in LUS images. Using the dual-tree complex wavelet transform (DTCWT) and four types of common image features we propose a method to classify the LUS images into 6 common neonatal lung conditions. These conditions are normal lung, pneumothorax (PTX), transient tachypnea of the newborn (TTN), respiratory distress syndrome (RDS), chronic lung disease (CLD) and consolidation (CON) that could be pneumonia or atelectasis. The proposed method using DTCWT decomposition extracted global statistical, grey-level co-occurrence matrix (GLCM), grey-level run length matrix (GLRLM) and linear binary pattern (LBP) features to be fed to a linear discriminative analysis (LDA) based classifier. Using 15 best DTCWT features along with 3 clinical features the proposed approach achieved a per-image classification accuracy of 92.78% with a balanced dataset containing 720 images from 24 patients and 74.39% with the larger unbalanced dataset containing 1550 images from 42 patients. Likewise, the proposed method achieved a maximum per-subject classification accuracy of 81.53% with 43 DTCWT features and 3 clinical features using the balanced dataset and 64.97% with 13 DTCWT features and 3 clinical features using the unbalanced dataset.
COVID-19 (Coronavirus disease 2019) has been quickly spreading since its outbreak, impacting financial markets and healthcare systems globally. Countries all around the world have adopted a number of extraordinary steps to restrict the spreading virus, where early COVID-19 diagnosis is essential. Medical images such as X-ray images and Computed Tomography scans are becoming one of the main diagnostic tools to combat COVID-19 with the aid of deep learning-based systems. In this survey, we investigate the main contributions of deep learning applications using medical images in fighting against COVID-19 from the aspects of image classification, lesion localization, and severity quantification, and review different deep learning architectures and some image preprocessing techniques for achieving a preciser diagnosis. We also provide a summary of the X-ray and CT image datasets used in various studies for COVID-19 detection. The key difficulties and potential applications of deep learning in fighting against COVID-19 are finally discussed. This work summarizes the latest methods of deep learning using medical images to diagnose COVID-19, highlighting the challenges and inspiring more studies to keep utilizing the advantages of deep learning to combat COVID-19.
Training deep neural networks(DNN) with noisy labels is challenging since DNN can easily memorize inaccurate labels, leading to poor generalization ability. Recently, the meta-learning based label correction strategy is widely adopted to tackle this problem via identifying and correcting potential noisy labels with the help of a small set of clean validation data. Although training with purified labels can effectively improve performance, solving the meta-learning problem inevitably involves a nested loop of bi-level optimization between model weights and hyper-parameters (i.e., label distribution). As compromise, previous methods resort to a coupled learning process with alternating update. In this paper, we empirically find such simultaneous optimization over both model weights and label distribution can not achieve an optimal routine, consequently limiting the representation ability of backbone and accuracy of corrected labels. From this observation, a novel multi-stage label purifier named DMLP is proposed. DMLP decouples the label correction process into label-free representation learning and a simple meta label purifier. In this way, DMLP can focus on extracting discriminative feature and label correction in two distinctive stages. DMLP is a plug-and-play label purifier, the purified labels can be directly reused in naive end-to-end network retraining or other robust learning methods, where state-of-the-art results are obtained on several synthetic and real-world noisy datasets, especially under high noise levels.
Collecting large-scale datasets is crucial for training deep models, annotating the data, however, inevitably yields noisy labels, which poses challenges to deep learning algorithms. Previous efforts tend to mitigate this problem via identifying and removing noisy samples or correcting their labels according to the statistical properties (e.g., loss values) among training samples. In this paper, we aim to tackle this problem from a new perspective, delving into the deep feature maps, we empirically find that models trained with clean and mislabeled samples manifest distinguishable activation feature distributions. From this observation, a novel robust training approach termed adversarial noisy masking is proposed. The idea is to regularize deep features with a label quality guided masking scheme, which adaptively modulates the input data and label simultaneously, preventing the model to overfit noisy samples. Further, an auxiliary task is designed to reconstruct input data, it naturally provides noise-free self-supervised signals to reinforce the generalization ability of deep models. The proposed method is simple and flexible, it is tested on both synthetic and real-world noisy datasets, where significant improvements are achieved over previous state-of-the-art methods.
State space models (SSMs) have high performance on long sequence modeling but require sophisticated initialization techniques and specialized implementations for high quality and runtime performance. We study whether a simple alternative can match SSMs in performance and efficiency: directly learning long convolutions over the sequence. We find that a key requirement to achieving high performance is keeping the convolution kernels smooth. We find that simple interventions--such as squashing the kernel weights--result in smooth kernels and recover SSM performance on a range of tasks including the long range arena, image classification, language modeling, and brain data modeling. Next, we develop FlashButterfly, an IO-aware algorithm to improve the runtime performance of long convolutions. FlashButterfly appeals to classic Butterfly decompositions of the convolution to reduce GPU memory IO and increase FLOP utilization. FlashButterfly speeds up convolutions by 2.2$\times$, and allows us to train on Path256, a challenging task with sequence length 64K, where we set state-of-the-art by 29.1 points while training 7.2$\times$ faster than prior work. Lastly, we introduce an extension to FlashButterfly that learns the coefficients of the Butterfly decomposition, increasing expressivity without increasing runtime. Using this extension, we outperform a Transformer on WikiText103 by 0.2 PPL with 30% fewer parameters.
Computer aided diagnosis (CAD) has gained an increased amount of attention in the general research community over the last years as an example of a typical limited data application - with experiments on labeled 100k-200k datasets. Although these datasets are still small compared to natural image datasets like ImageNet1k, ImageNet21k and JFT, they are large for annotated medical datasets, where 1k-10k labeled samples are much more common. There is no baseline on which methods to build on in the low data regime. In this work we bridge this gap by providing an extensive study on medical image classification with limited annotations (5k). We present a study of modern architectures applied to a fixed low data regime of 5000 images on the CheXpert dataset. Conclusively we find that models pretrained on ImageNet21k achieve a higher AUC and larger models require less training steps. All models are quite well calibrated even though we only fine-tuned on 5000 training samples. All 'modern' architectures have higher AUC than ResNet50. Regularization of Big Transfer Models with MixUp or Mean Teacher improves calibration, MixUp also improves accuracy. Vision Transformer achieve comparable or on par results to Big Transfer Models.
Deep neural networks (DNNs) and, in particular, convolutional neural networks (CNNs) have brought significant advances in a wide range of modern computer application problems. However, the increasing availability of large amounts of datasets as well as the increasing available computational power of modern computers lead to a steady growth in the complexity and size of DNN and CNN models, respectively, and thus, to longer training times. Hence, various methods and attempts have been developed to accelerate and parallelize the training of complex network architectures. In this work, a novel CNN-DNN architecture is proposed that naturally supports a model parallel training strategy and that is loosely inspired by two-level domain decomposition methods (DDM). First, local CNN models, that is, subnetworks, are defined that operate on overlapping or nonoverlapping parts of the input data, for example, sub-images. The subnetworks can be trained completely in parallel and independently of each other. Each subnetwork then outputs a local decision for the given machine learning problem which is exclusively based on the respective local input data. Subsequently, in a second step, an additional DNN model is trained which evaluates the local decisions of the local subnetworks and generates a final, global decision. In this paper, we apply the proposed architecture to image classification problems using CNNs. Experimental results for different 2D image classification problems are provided as well as a face recognition problem, and a classification problem for 3D computer tomography (CT) scans. Therefore, classical ResNet and VGG architectures are considered. The results show that the proposed approach can significantly accelerate the required training time compared to the global model and, additionally, can also help to improve the accuracy of the underlying classification problem.
Semantic image segmentation (SiS) plays a fundamental role in a broad variety of computer vision applications, providing key information for the global understanding of an image. This survey is an effort to summarize two decades of research in the field of SiS, where we propose a literature review of solutions starting from early historical methods followed by an overview of more recent deep learning methods including the latest trend of using transformers. We complement the review by discussing particular cases of the weak supervision and side machine learning techniques that can be used to improve the semantic segmentation such as curriculum, incremental or self-supervised learning. State-of-the-art SiS models rely on a large amount of annotated samples, which are more expensive to obtain than labels for tasks such as image classification. Since unlabeled data is instead significantly cheaper to obtain, it is not surprising that Unsupervised Domain Adaptation (UDA) reached a broad success within the semantic segmentation community. Therefore, a second core contribution of this book is to summarize five years of a rapidly growing field, Domain Adaptation for Semantic Image Segmentation (DASiS) which embraces the importance of semantic segmentation itself and a critical need of adapting segmentation models to new environments. In addition to providing a comprehensive survey on DASiS techniques, we unveil also newer trends such as multi-domain learning, domain generalization, domain incremental learning, test-time adaptation and source-free domain adaptation. Finally, we conclude this survey by describing datasets and benchmarks most widely used in SiS and DASiS and briefly discuss related tasks such as instance and panoptic image segmentation, as well as applications such as medical image segmentation.
Deep neural networks (DNNs) have demonstrated remarkable performance across various tasks, including image and speech recognition. However, maximizing the effectiveness of DNNs requires meticulous optimization of numerous hyperparameters and network parameters through training. Moreover, high-performance DNNs entail many parameters, which consume significant energy during training. In order to overcome these challenges, researchers have turned to spiking neural networks (SNNs), which offer enhanced energy efficiency and biologically plausible data processing capabilities, rendering them highly suitable for sensory data tasks, particularly in neuromorphic data. Despite their advantages, SNNs, like DNNs, are susceptible to various threats, including adversarial examples and backdoor attacks. Yet, the field of SNNs still needs to be explored in terms of understanding and countering these attacks. This paper delves into backdoor attacks in SNNs using neuromorphic datasets and diverse triggers. Specifically, we explore backdoor triggers within neuromorphic data that can manipulate their position and color, providing a broader scope of possibilities than conventional triggers in domains like images. We present various attack strategies, achieving an attack success rate of up to 100% while maintaining a negligible impact on clean accuracy. Furthermore, we assess these attacks' stealthiness, revealing that our most potent attacks possess significant stealth capabilities. Lastly, we adapt several state-of-the-art defenses from the image domain, evaluating their efficacy on neuromorphic data and uncovering instances where they fall short, leading to compromised performance.
For the stable optimization of deep neural networks, regularization methods such as dropout and batch normalization have been used in various tasks. Nevertheless, the correct position to apply dropout has rarely been discussed, and different positions have been employed depending on the practitioners. In this study, we investigate the correct position to apply dropout. We demonstrate that for a residual network with batch normalization, applying dropout at certain positions increases the performance, whereas applying dropout at other positions decreases the performance. Based on theoretical analysis, we provide the following guideline for the correct position to apply dropout: apply one dropout after the last batch normalization but before the last weight layer in the residual branch. We provide detailed theoretical explanations to support this claim and demonstrate them through module tests. In addition, we investigate the correct position of dropout in the head that produces the final prediction. Although the current consensus is to apply dropout after global average pooling, we prove that applying dropout before global average pooling leads to a more stable output. The proposed guidelines are validated through experiments using different datasets and models.
The public model zoo containing enormous powerful pretrained model families (e.g., ResNet/DeiT) has reached an unprecedented scope than ever, which significantly contributes to the success of deep learning. As each model family consists of pretrained models with diverse scales (e.g., DeiT-Ti/S/B), it naturally arises a fundamental question of how to efficiently assemble these readily available models in a family for dynamic accuracy-efficiency trade-offs at runtime. To this end, we present Stitchable Neural Networks (SN-Net), a novel scalable and efficient framework for model deployment. It cheaply produces numerous networks with different complexity and performance trade-offs given a family of pretrained neural networks, which we call anchors. Specifically, SN-Net splits the anchors across the blocks/layers and then stitches them together with simple stitching layers to map the activations from one anchor to another. With only a few epochs of training, SN-Net effectively interpolates between the performance of anchors with varying scales. At runtime, SN-Net can instantly adapt to dynamic resource constraints by switching the stitching positions. Extensive experiments on ImageNet classification demonstrate that SN-Net can obtain on-par or even better performance than many individually trained networks while supporting diverse deployment scenarios. For example, by stitching Swin Transformers, we challenge hundreds of models in Timm model zoo with a single network. We believe this new elastic model framework can serve as a strong baseline for further research in wider communities.
Ever since their conception, Transformers have taken over traditional sequence models in many tasks, such as NLP, image classification, and video/audio processing, for their fast training and superior performance. Much of the merit is attributable to positional encoding and multi-head attention. However, Transformers fall short in learning long-range dependencies mainly due to the quadratic complexity scaled with context length, in terms of both time and space. Consequently, over the past five years, a myriad of methods has been proposed to make Transformers more efficient. In this work, we first take a step back, study and compare existing solutions to long-sequence modeling in terms of their pure mathematical formulation. Specifically, we summarize them using a unified template, given their shared nature of token mixing. Through benchmarks, we then demonstrate that long context length does yield better performance, albeit application-dependent, and traditional Transformer models fall short in taking advantage of long-range dependencies. Next, inspired by emerging sparse models of huge capacity, we propose a machine learning system for handling million-scale dependencies. As a proof of concept, we evaluate the performance of one essential component of this system, namely, the distributed multi-head attention. We show that our algorithm can scale up attention computation by almost $40\times$ using four GeForce RTX 4090 GPUs, compared to vanilla multi-head attention mechanism. We believe this study is an instrumental step towards modeling million-scale dependencies.
Renal cell carcinoma (RCC) is a type of cancer that originates in the kidneys and is the most common type of kidney cancer in adults. It can be classified into several subtypes, including clear cell RCC, papillary RCC, and chromophobe RCC. In this study, an artificial intelligence model was developed and trained for classifying different subtypes of RCC using ResNet-18, a convolutional neural network that has been widely used for image classification tasks. The model was trained on a dataset of RCC histopathology images, which consisted of digital images of RCC surgical resection slides that were annotated with the corresponding subtype labels. The performance of the trained model was evaluated using several metrics, including accuracy, precision, and recall. Additionally, in this research, a novel synthetic image generation tool, NephroNet, is developed on diffusion models that are used to generate original images of RCC surgical resection slides. Diffusion models are a class of generative models capable of synthesizing high-quality images from noise. Several diffusers such as Stable Diffusion, Dreambooth Text-to-Image, and Textual Inversion were trained on a dataset of RCC images and were used to generate a series of original images that resembled RCC surgical resection slides, all within the span of fewer than four seconds. The generated images were visually realistic and could be used for creating new training datasets, testing the performance of image analysis algorithms, and training medical professionals. NephroNet is provided as an open-source software package and contains files for data preprocessing, training, and visualization. Overall, this study demonstrates the potential of artificial intelligence and diffusion models for classifying and generating RCC images, respectively. These methods could be useful for improving the diagnosis and treatment of RCC and more.
Non-linear activation functions are one of the main parts of deep neural network architectures. The choice of the activation function can affect model speed, performance and convergence. Most popular activation functions don't have any trainable parameters and don't alter during the training. We propose different activation functions with and without trainable parameters. Said activation functions have a number of advantages and disadvantages. We'll be testing the performance of said activation functions and comparing the results with widely known activation function ReLU. We assume that the activation functions with trainable parameters can outperform functions without ones, because the trainable parameters allow the model to "select'' the type of each of the activation functions itself, however, this strongly depends on the architecture of the deep neural network and the activation function itself.
We present a synchronized multisensory dataset for accurate and robust indoor localization: the Lund University Vision, Radio, and Audio (LuViRA) Dataset. The dataset includes color images, corresponding depth maps, inertial measurement unit (IMU) readings, channel response between a 5G massive multiple-input and multiple-output (MIMO) testbed and user equipment, audio recorded by 12 microphones, and accurate six degrees of freedom (6DOF) pose ground truth of 0.5 mm. We synchronize these sensors to ensure that all data is recorded simultaneously. A camera, speaker, and transmit antenna are placed on top of a slowly moving service robot, and 89 trajectories are recorded. Each trajectory includes 20 to 50 seconds of recorded sensor data and ground truth labels. Data from different sensors can be used separately or jointly to perform localization tasks, and data from the motion capture (mocap) system is used to verify the results obtained by the localization algorithms. The main aim of this dataset is to enable research on sensor fusion with the most commonly used sensors for localization tasks. Moreover, the full dataset or some parts of it can also be used for other research areas such as channel estimation, image classification, etc. Our dataset is available at: https://github.com/ilaydayaman/LuViRA_Dataset
Deep learning methods have been successfully applied to hyperspectral image (HSI) classification with remarkable performance. Because of limited labelled HSI data, earlier studies primarily adopted a patch-based classification framework, which divides images into overlapping patches for training and testing. However, this approach results in redundant computations and possible information leakage. In this study, we propose an objective evaluation-based high-efficiency learning framework for tiny HSI classification. This framework comprises two main parts: (i) a leakage-free balanced sampling strategy, and (ii) a modified end-to-end fully convolutional network (FCN) architecture that optimizes the trade-off between accuracy and efficiency. The leakage-free balanced sampling strategy generates balanced and non-overlapping training and testing data by partitioning an HSI and the ground truth image into small windows, each of which corresponds to one training or testing sample. The proposed high-efficiency FCN exhibits a pixel-to-pixel architecture with modifications aimed at faster inference speed and improved parameter efficiency. Experiments conducted on four representative datasets demonstrated that the proposed sampling strategy can provide objective performance evaluation and that the proposed network outperformed many state-of-the-art approaches with respect to the speed/accuracy tradeoff. Our source code is available at https://github.com/xmzhang2018.
A text on an image often stores important information and directly carries high level semantics, makes it as important source of information and become a very active research topic. Many studies have shown that the use of CNN-based neural networks is quite effective and accurate for image classification which is the basis of text recognition. It can also be more enhanced by using transfer learning from pre-trained model trained on ImageNet dataset as an initial weight. In this research, the recognition is trained by using Chars74K dataset and the best model results then tested on some samples of IIIT-5K-Dataset. The research results showed that the best accuracy is the model that trained using VGG-16 architecture applied with image transformation of rotation 15{\deg}, image scale of 0.9, and the application of gaussian blur effect. The research model has an accuracy of 97.94% for validation data, 98.16% for test data, and 95.62% for the test data from IIIT-5K-Dataset. Based on these results, it can be concluded that pre-trained CNN can produce good accuracy for text recognition, and the model architecture that used in this study can be used as reference material in the development of text detection systems in the future
Contextual information plays an important role in many computer vision tasks, such as object detection, video action detection, image classification, etc. Recognizing a single object or action out of context could be sometimes very challenging, and context information may help improve the understanding of a scene or an event greatly. Appearance context information, e.g., colors or shapes of the background of an object can improve the recognition accuracy of the object in the scene. Semantic context (e.g. a keyboard on an empty desk vs. a keyboard next to a desktop computer ) will improve accuracy and exclude unrelated events. Context information that are not in the image itself, such as the time or location of an images captured, can also help to decide whether certain event or action should occur. Other types of context (e.g. 3D structure of a building) will also provide additional information to improve the accuracy. In this survey, different context information that has been used in computer vision tasks is reviewed. We categorize context into different types and different levels. We also review available machine learning models and image/video datasets that can employ context information. Furthermore, we compare context based integration and context-free integration in mainly two classes of tasks: image-based and video-based. Finally, this survey is concluded by a set of promising future directions in context learning and utilization.
The scaling of Transformers has driven breakthrough capabilities for language models. At present, the largest large language models (LLMs) contain upwards of 100B parameters. Vision Transformers (ViT) have introduced the same architecture to image and video modelling, but these have not yet been successfully scaled to nearly the same degree; the largest dense ViT contains 4B parameters (Chen et al., 2022). We present a recipe for highly efficient and stable training of a 22B-parameter ViT (ViT-22B) and perform a wide variety of experiments on the resulting model. When evaluated on downstream tasks (often with a lightweight linear model on frozen features), ViT-22B demonstrates increasing performance with scale. We further observe other interesting benefits of scale, including an improved tradeoff between fairness and performance, state-of-the-art alignment to human visual perception in terms of shape/texture bias, and improved robustness. ViT-22B demonstrates the potential for "LLM-like" scaling in vision, and provides key steps towards getting there.