We consider the problem of explaining the decisions of deep neural networks for image recognition in terms of human-recognizable visual concepts. In particular, given a test set of images, we aim to explain each classification in terms of a small number of image regions, or activation maps, which have been associated with semantic concepts by a human annotator. This allows for generating summary views of the typical reasons for classifications, which can help build trust in a classifier and/or identify example types for which the classifier may not be trusted. For this purpose, we developed a user interface for "interactive naming," which allows a human annotator to manually cluster significant activation maps in a test set into meaningful groups called "visual concepts". The main contribution of this paper is a systematic study of the visual concepts produced by five human annotators using the interactive naming interface. In particular, we consider the adequacy of the concepts for explaining the classification of test-set images, correspondence of the concepts to activations of individual neurons, and the inter-annotator agreement of visual concepts. We find that a large fraction of the activation maps have recognizable visual concepts, and that there is significant agreement between the different annotators about their denotations. Our work is an exploratory study of the interplay between machine learning and human recognition mediated by visualizations of the results of learning.
Few-shot image classification requires the classifier to robustly cope with unseen classes even if there are only a few samples for each class. Recent advances benefit from the meta-learning process where episodic tasks are formed to train a model that can adapt to class change. However, these task sare independent to each other and existing works mainly rely on limited samples of individual support set in a single meta task. This strategy leads to severe meta shift issues across multiple tasks, meaning the learned prototypes or class descriptors are not stable as each task only involves their own support set. To avoid this problem, we propose a concise Semantic RegularizationNetwork to learn a common semantic space under the framework of meta-learning. In this space, all class descriptors can be regularized by the learned semantic basis, which can effectively solve the meta shift problem. The key is to train a class encoder and decoder structure that can encode the sample embedding features into the semantic domain with trained semantic basis, and generate a more stable and general class descriptor from the decoder. We evaluate our work by extensive comparisons with previous methods on three benchmark datasets (MiniImageNet, TieredImageNet, and CUB). The results show that the semantic regularization module improves performance by 4%-7% over the baseline method, and achieves competitive results over the current state-of-the-art models.
Optimization techniques are of great importance to effectively and efficiently train a deep neural network (DNN). It has been shown that using the first and second order statistics (e.g., mean and variance) to perform Z-score standardization on network activations or weight vectors, such as batch normalization (BN) and weight standardization (WS), can improve the training performance. Different from these existing methods that mostly operate on activations or weights, we present a new optimization technique, namely gradient centralization (GC), which operates directly on gradients by centralizing the gradient vectors to have zero mean. GC can be viewed as a projected gradient descent method with a constrained loss function. We show that GC can regularize both the weight space and output feature space so that it can boost the generalization performance of DNNs. Moreover, GC improves the Lipschitzness of the loss function and its gradient so that the training process becomes more efficient and stable. GC is very simple to implement and can be easily embedded into existing gradient based DNN optimizers with only one line of code. It can also be directly used to fine-tune the pre-trained DNNs. Our experiments on various applications, including general image classification, fine-grained image classification, detection and segmentation, demonstrate that GC can consistently improve the performance of DNN learning. The code of GC can be found at https://github.com/Yonghongwei/Gradient-Centralization.
This paper focuses on webly supervised learning (WSL), where datasets are built by crawling samples from the Internet and directly using search queries as web labels. Although WSL benefits from fast and low-cost data collection, noises in web labels hinder better performance of the image classification model. To alleviate this problem, in recent works, self-label supervised loss $\mathcal{L}_s$ is utilized together with webly supervised loss $\mathcal{L}_w$. $\mathcal{L}_s$ relies on pseudo labels predicted by the model itself. Since the correctness of the web label or pseudo label is usually on a case-by-case basis for each web sample, it is desirable to adjust the balance between $\mathcal{L}_s$ and $\mathcal{L}_w$ on sample level. Inspired by the ability of Deep Neural Networks (DNNs) in confidence prediction, we introduce Self-Contained Confidence (SCC) by adapting model uncertainty for WSL setting, and use it to sample-wisely balance $\mathcal{L}_s$ and $\mathcal{L}_w$. Therefore, a simple yet effective WSL framework is proposed. A series of SCC-friendly regularization approaches are investigated, among which the proposed graph-enhanced mixup is the most effective method to provide high-quality confidence to enhance our framework. The proposed WSL framework has achieved the state-of-the-art results on two large-scale WSL datasets, WebVision-1000 and Food101-N. Code is available at https://github.com/bigvideoresearch/SCC.
Visual scene understanding is the core task in making any crucial decision in any computer vision system. Although popular computer vision datasets like Cityscapes, MS-COCO, PASCAL provide good benchmarks for several tasks (e.g. image classification, segmentation, object detection), these datasets are hardly suitable for post disaster damage assessments. On the other hand, existing natural disaster datasets include mainly satellite imagery which have low spatial resolution and a high revisit period. Therefore, they do not have a scope to provide quick and efficient damage assessment tasks. Unmanned Aerial Vehicle(UAV) can effortlessly access difficult places during any disaster and collect high resolution imagery that is required for aforementioned tasks of computer vision. To address these issues we present a high resolution UAV imagery, FloodNet, captured after the hurricane Harvey. This dataset demonstrates the post flooded damages of the affected areas. The images are labeled pixel-wise for semantic segmentation task and questions are produced for the task of visual question answering. FloodNet poses several challenges including detection of flooded roads and buildings and distinguishing between natural water and flooded water. With the advancement of deep learning algorithms, we can analyze the impact of any disaster which can make a precise understanding of the affected areas. In this paper, we compare and contrast the performances of baseline methods for image classification, semantic segmentation, and visual question answering on our dataset.
Hyperspectral images (HSI) not only have a broad macroscopic field of view but also contain rich spectral information, and the types of surface objects can be identified through spectral information, which is one of the main applications in hyperspectral image related research.In recent years, more and more deep learning methods have been proposed, among which convolutional neural networks (CNN) are the most influential. However, CNN-based methods are difficult to capture long-range dependencies, and also require a large amount of labeled data for model training.Besides, most of the self-supervised training methods in the field of HSI classification are based on the reconstruction of input samples, and it is difficult to achieve effective use of unlabeled samples. To address the shortcomings of CNN networks, we propose a noval multi-scale convolutional embedding module for HSI to realize effective extraction of spatial-spectral information, which can be better combined with Transformer network.In order to make more efficient use of unlabeled data, we propose a new self-supervised pretask. Similar to Mask autoencoder, but our pre-training method only masks the corresponding token of the central pixel in the encoder, and inputs the remaining token into the decoder to reconstruct the spectral information of the central pixel.Such a pretask can better model the relationship between the central feature and the domain feature, and obtain more stable training results.
As the technology industry is moving towards implementing tasks such as natural language processing, path planning, image classification, and more on smaller edge computing devices, the demand for more efficient implementations of algorithms and hardware accelerators has become a significant area of research. In recent years, several edge deep learning hardware accelerators have been released that specifically focus on reducing the power and area consumed by deep neural networks (DNNs). On the other hand, spiking neural networks (SNNs) which operate on discrete time-series data, have been shown to achieve substantial power reductions over even the aforementioned edge DNN accelerators when deployed on specialized neuromorphic event-based/asynchronous hardware. While neuromorphic hardware has demonstrated great potential for accelerating deep learning tasks at the edge, the current space of algorithms and hardware is limited and still in rather early development. Thus, many hybrid approaches have been proposed which aim to convert pre-trained DNNs into SNNs. In this work, we provide a general guide to converting pre-trained DNNs into SNNs while also presenting techniques to improve the deployment of converted SNNs on neuromorphic hardware with respect to latency, power, and energy. Our experimental results show that when compared against the Intel Neural Compute Stick 2, Intel's neuromorphic processor, Loihi, consumes up to 27x less power and 5x less energy in the tested image classification tasks by using our SNN improvement techniques.
The weight-sharing mechanism of convolutional kernels ensures translation-equivariance of convolution neural networks (CNNs). Recently, rotation-equivariance has been investigated. However, research on scale-equivariance or simultaneous scale-rotation-equivariance is insufficient. This study proposes a Lie group-CNN, which can keep scale-rotation-equivariance for image classification tasks. The Lie group-CNN includes a lifting module, a series of group convolution modules, a global pooling layer, and a classification layer. The lifting module transfers the input image from Euclidean space to Lie group space, and the group convolution is parameterized through a fully connected network using Lie-algebra of Lie-group elements as inputs to achieve scale-rotation-equivariance. The Lie group SIM(2) is utilized to establish the Lie group-CNN with scale-rotation-equivariance. Scale-rotation-equivariance of Lie group-CNN is verified and achieves the best recognition accuracy on the blood cell dataset (97.50%) and the HAM10000 dataset (77.90%) superior to Lie algebra convolution network, dilation convolution, spatial transformer network, and scale-equivariant steerable network. In addition, the generalization ability of the Lie group-CNN on SIM(2) on rotation-equivariance is verified on rotated-MNIST and rotated-CIFAR10, and the robustness of the network is verified on SO(2) and SE(2). Therefore, the Lie group-CNN can successfully extract geometric features and performs equivariant recognition on images with rotation and scale transformations.
Few-shot image classification aims to accurately classify unlabeled images using only a few labeled samples. The state-of-the-art solutions are built by deep learning, which focuses on designing increasingly complex deep backbones. Unfortunately, the task remains very challenging due to the difficulty of transferring the knowledge learned in training classes to new ones. In this paper, we propose a novel approach based on the non-i.i.d paradigm of gradual machine learning (GML). It begins with only a few labeled observations, and then gradually labels target images in the increasing order of hardness by iterative factor inference in a factor graph. Specifically, our proposed solution extracts indicative feature representations by deep backbones, and then constructs both unary and binary factors based on the extracted features to facilitate gradual learning. The unary factors are constructed based on class center distance in an embedding space, while the binary factors are constructed based on k-nearest neighborhood. We have empirically validated the performance of the proposed approach on benchmark datasets by a comparative study. Our extensive experiments demonstrate that the proposed approach can improve the SOTA performance by 1-5% in terms of accuracy. More notably, it is more robust than the existing deep models in that its performance can consistently improve as the size of query set increases while the performance of deep models remains essentially flat or even becomes worse.
Recent advances in training deep (multi-layer) architectures have inspired a renaissance in neural network use. For example, deep convolutional networks are becoming the default option for difficult tasks on large datasets, such as image and speech recognition. However, here we show that error rates below 1% on the MNIST handwritten digit benchmark can be replicated with shallow non-convolutional neural networks. This is achieved by training such networks using the 'Extreme Learning Machine' (ELM) approach, which also enables a very rapid training time (~10 minutes). Adding distortions, as is common practise for MNIST, reduces error rates even further. Our methods are also shown to be capable of achieving less than 5.5% error rates on the NORB image database. To achieve these results, we introduce several enhancements to the standard ELM algorithm, which individually and in combination can significantly improve performance. The main innovation is to ensure each hidden-unit operates only on a randomly sized and positioned patch of each image. This form of random `receptive field' sampling of the input ensures the input weight matrix is sparse, with about 90% of weights equal to zero. Furthermore, combining our methods with a small number of iterations of a single-batch backpropagation method can significantly reduce the number of hidden-units required to achieve a particular performance. Our close to state-of-the-art results for MNIST and NORB suggest that the ease of use and accuracy of the ELM algorithm for designing a single-hidden-layer neural network classifier should cause it to be given greater consideration either as a standalone method for simpler problems, or as the final classification stage in deep neural networks applied to more difficult problems.
Skin cancer is the most common cancer in the existing world constituting one-third of the cancer cases. Benign skin cancers are not fatal, can be cured with proper medication. But it is not the same as the malignant skin cancers. In the case of malignant melanoma, in its peak stage, the maximum life expectancy is less than or equal to 5 years. But, it can be cured if detected in early stages. Though there are numerous clinical procedures, the accuracy of diagnosis falls between 49% to 81% and is time-consuming. So, dermoscopy has been brought into the picture. It helped in increasing the accuracy of diagnosis but could not demolish the error-prone behaviour. A quick and less error-prone solution is needed to diagnose this majorly growing skin cancer. This project deals with the usage of deep learning in skin lesion classification. In this project, an automated model for skin lesion classification using dermoscopic images has been developed with CNN(Convolution Neural Networks) as a training model. Convolution neural networks are known for capturing features of an image. So, they are preferred in analyzing medical images to find the characteristics that drive the model towards success. Techniques like data augmentation for tackling class imbalance, segmentation for focusing on the region of interest and 10-fold cross-validation to make the model robust have been brought into the picture. This project also includes usage of certain preprocessing techniques like brightening the images using piece-wise linear transformation function, grayscale conversion of the image, resize the image. This project throws a set of valuable insights on how the accuracy of the model hikes with the bringing of new input strategies, preprocessing techniques. The best accuracy this model could achieve is 0.886.
텍스트 정보를 활용한 이미지 클러스터링은 비전-언어 모델(VLM)이 학습한 강력한 표현 덕분에 유망한 연구 방향으로 부상했다. 기존 방법은 일반적으로 각 이미지에 대한 텍스트 대응물을 검색한 뒤, 사전학습된 VLM에서 이어받은 이미지-텍스트 유사도 최대화 등과 같이 교차 모달 일치를 직접 강제함으로써 다중 모달 표현을 정교화한다. 그러나 이러한 전략은 각 모달리티 내부의 고유 구조를 명시적으로 모델링하지 않은 채 이질적 표현을 모달리티 간에 정렬하므로, 신뢰할 수 없는 정렬을 초래하거나 클러스터링에 중요한 모달리티별 구조를 왜곡할 수 있다. 본 논문에서는 모달리티 공유 자기표현 모델을 통해 교차 모달 구조를 발견하고, 동시에 모달리티별 부분공간들의 합집합에 부합하는 구조화된 표현을 학습하는 단순하지만 원칙적인 접근법인 심층 모달리티 공유 자기표현 모델(DeepMORSE)을 제안한다. 또한 모달리티 공유 자기표현 계수가 클래스 간 노이즈를 억제하여 부분공간 보존(subspace-preserving) 해로 향함을 이론적으로 정당화하고, 미니배치 최적화 절차가 자기표현 모델에 암묵적 정규화를 도입함을 보인다. DeepMORSE를 널리 사용되는 6개 이미지 클러스터링 벤치마크에서 평가한 결과, UCF-101, DTD-47, ImageNet-Dogs 데이터셋에서 3%를 넘는 성능 향상을 관찰했다. 또한 작업별 손실이나 후처리 없이도 이미지 검색 및 제로샷 분류 같은 다운스트림 작업에서 state-of-the-art 성능을 달성함으로써 학습된 표현의 강한 전이 가능성을 입증했다. 코드는 https://github.com/mengxianghan123/DeepMORSE 에서 제공된다.
In biomedical image analysis, the applicability of deep learning methods is directly impacted by the quantity of image data available. This is due to deep learning models requiring large image datasets to provide high-level performance. Generative Adversarial Networks (GANs) have been widely utilized to address data limitations through the generation of synthetic biomedical images. GANs consist of two models. The generator, a model that learns how to produce synthetic images based on the feedback it receives. The discriminator, a model that classifies an image as synthetic or real and provides feedback to the generator. Throughout the training process, a GAN can experience several technical challenges that impede the generation of suitable synthetic imagery. First, the mode collapse problem whereby the generator either produces an identical image or produces a uniform image from distinct input features. Second, the non-convergence problem whereby the gradient descent optimizer fails to reach a Nash equilibrium. Thirdly, the vanishing gradient problem whereby unstable training behavior occurs due to the discriminator achieving optimal classification performance resulting in no meaningful feedback being provided to the generator. These problems result in the production of synthetic imagery that is blurry, unrealistic, and less diverse. To date, there has been no survey article outlining the impact of these technical challenges in the context of the biomedical imagery domain. This work presents a review and taxonomy based on solutions to the training problems of GANs in the biomedical imaging domain. This survey highlights important challenges and outlines future research directions about the training of GANs in the domain of biomedical imagery.
Brain imaging analysis on clinically acquired computed tomography (CT) is essential for the diagnosis, risk prediction of progression, and treatment of the structural phenotypes of traumatic brain injury (TBI). However, in real clinical imaging scenarios, entire body CT images (e.g., neck, abdomen, chest, pelvis) are typically captured along with whole brain CT scans. For instance, in a typical sample of clinical TBI imaging cohort, only ~15% of CT scans actually contain whole brain CT images suitable for volumetric brain analyses; the remaining are partial brain or non-brain images. Therefore, a manual image retrieval process is typically required to isolate the whole brain CT scans from the entire cohort. However, the manual image retrieval is time and resource consuming and even more difficult for the larger cohorts. To alleviate the manual efforts, in this paper we propose an automated 3D medical image retrieval pipeline, called deep montage-based image retrieval (dMIR), which performs classification on 2D montage images via a deep convolutional neural network. The novelty of the proposed method for image processing is to characterize the medical image retrieval task based on the montage images. In a cohort of 2000 clinically acquired TBI scans, 794 scans were used as training data, 206 scans were used as validation data, and the remaining 1000 scans were used as testing data. The proposed achieved accuracy=1.0, recall=1.0, precision=1.0, f1=1.0 for validation data, while achieved accuracy=0.988, recall=0.962, precision=0.962, f1=0.962 for testing data. Thus, the proposed dMIR is able to perform accurate CT whole brain image retrieval from large-scale clinical cohorts.
This paper considers attacks against machine learning algorithms used in remote sensing applications, a domain that presents a suite of challenges that are not fully addressed by current research focused on natural image data such as ImageNet. In particular, we present a new study of adversarial examples in the context of satellite image classification problems. Using a recently curated data set and associated classifier, we provide a preliminary analysis of adversarial examples in settings where the targeted classifier is permitted multiple observations of the same location over time. While our experiments to date are purely digital, our problem setup explicitly incorporates a number of practical considerations that a real-world attacker would need to take into account when mounting a physical attack. We hope this work provides a useful starting point for future studies of potential vulnerabilities in this setting.
Although traditionally binary visual representations are mainly designed to reduce computational and storage costs in the image retrieval research, this paper argues that binary visual representations can be applied to large scale recognition and detection problems in addition to hashing in retrieval. Furthermore, the binary nature may make it generalize better than its real-valued counterparts. Existing binary hashing methods are either two-stage or hinging on loss term regularization or saturated functions, hence converge slowly and only emit soft binary values. This paper proposes Approximately Binary Clamping (ABC), which is non-saturating, end-to-end trainable, with fast convergence and can output true binary visual representations. ABC achieves comparable accuracy in ImageNet classification as its real-valued counterpart, and even generalizes better in object detection. On benchmark image retrieval datasets, ABC also outperforms existing hashing methods.
A novel and fast semi-automatic method for segmentation, locating and counting blood cells in an image is proposed. In this method, thresholding is used to separate the nucleus from the other parts. We also use Hough transform for circles to locate the center of white cells. Locating and counting of red cells is performed using template matching. We make use of finding local maxima, labeling and mean value computation in order to shrink the areas obtained after applying Hough transform or template matching, to a single pixel as representative of location of each region. The proposed method is very fast and computes the number and location of white cells accurately. It is also capable of locating and counting the red cells with a small error.
In this paper, we present an ultra lightweight system that can effectively recognize different circuit components in an image with very limited training data. Along with the system, we also release the data set we created for the task. A two-stage approach is employed by our system. Selective search was applied to find the location of each circuit component. Based on its result, we crop the original image into smaller pieces. The pieces are then fed to the Convolutional Neural Network (CNN) for classification to identify each circuit component. It is of engineering significance and works well in circuit component recognition in a low resource setting. The accuracy of our system reaches 93.4\%, outperforming the support vector machine (SVM) baseline (75.00%) and the existing state-of-the-art RetinaNet solutions (92.80%).
Deep learning is the state-of-the-art for medical imaging tasks, but requires large, labeled datasets. For risk prediction, large datasets are rare since they require both imaging and follow-up (e.g., diagnosis codes). However, the release of publicly available imaging data with diagnostic labels presents an opportunity for self and semi-supervised approaches to improve label efficiency for risk prediction. Though several studies have compared self-supervised approaches in natural image classification, object detection, and medical image interpretation, there is limited data on which approaches learn robust representations for risk prediction. We present a comparison of semi- and self-supervised learning to predict mortality risk using chest x-ray images. We find that a semi-supervised autoencoder outperforms contrastive and transfer learning in internal and external validation.
In the realm of face image quality assesment (FIQA), method based on sample relative classification have shown impressive performance. However, the quality scores used as pseudo-labels assigned from images of classes with low intra-class variance could be unrelated to the actual quality in this method. To address this issue, we present IG-FIQA, a novel approach to guide FIQA training, introducing a weight parameter to alleviate the adverse impact of these classes. This method involves estimating sample intra-class variance at each iteration during training, ensuring minimal computational overhead and straightforward implementation. Furthermore, this paper proposes an on-the-fly data augmentation methodology for improved generalization performance in FIQA. On various benchmark datasets, our proposed method, IG-FIQA, achieved novel state-of-the-art (SOTA) performance.