Despite the recent success of deep neural networks, it remains challenging to effectively model the long-tail class distribution in visual recognition tasks. To address this problem, we first investigate the performance bottleneck of the two-stage learning framework via ablative study. Motivated by our discovery, we propose a unified distribution alignment strategy for long-tail visual recognition. Specifically, we develop an adaptive calibration function that enables us to adjust the classification scores for each data point. We then introduce a generalized re-weight method in the two-stage learning to balance the class prior, which provides a flexible and unified solution to diverse scenarios in visual recognition tasks. We validate our method by extensive experiments on four tasks, including image classification, semantic segmentation, object detection, and instance segmentation. Our approach achieves the state-of-the-art results across all four recognition tasks with a simple and unified framework. The code and models will be made publicly available at: https://github.com/Megvii-BaseDetection/DisAlign
We introduce a machine learning-based method for fully automated diagnosis of sickle cell disease of poor-quality unstained images of a mobile microscope. Our method is capable of distinguishing between diseased, trait (carrier), and normal samples unlike the previous methods that are limited to distinguishing the normal from the abnormal samples only. The novelty of this method comes from distinguishing the trait and the diseased samples from challenging images that have been captured directly in the field. The proposed approach contains two parts, the segmentation part followed by the classification part. We use a random forest algorithm to segment such challenging images acquitted through a mobile phone-based microscope. Then, we train two classifiers based on a random forest (RF) and a support vector machine (SVM) for classification. The results show superior performances of both of the classifiers not only for images which have been captured in the lab, but also for the ones which have been acquired in the field itself.
The prediction of a stock price has always been a challenging issue, as its volatility can be affected by many factors such as national policies, company financial reports, industry performance, and investor sentiment etc.. In this paper, we present a prediction model based on deep CNN and the candle charts, the continuous time stock information is processed. According to different information richness, prediction time interval and classification method, the original data is divided into multiple categories as the training set of CNN. In addition, the convolutional neural network is used to predict the stock market and analyze the difference in accuracy under different classification methods. The results show that the method has the best performance when the forecast time interval is 20 days. Moreover, the Moving Average Convergence Divergence and three kinds of moving average are added as input. This method can accurately predict the stock trend of the US NDAQ exchange for 92.2%. Meanwhile, this article distinguishes three conventional classification methods to provide guidance for future research.
Thermostability is an important prerequisite for enzymes employed for industrial applications. Several machine learning based models have thus been formulated for protein classification based on this particular trait. These models have employed features derived from sequences, structures or both resulting in a >93% accuracy based on a 10-fold cross-validation. Besides using various proteins from a wide range of organisms, such studies also rely on hundreds of features. In the present study, an enzyme specific classification model was created using significantly less number of features that provides a similar accuracy of classification for thermophilic and non-thermophilic enzyme serine proteases. For building the classifier, 219 thermophilic and 200 mesophilic bacterial genomes were mined for their respective serine protease sequences. Features were extracted for 800 sequences followed by feature selection. We deployed a random forest based classifier that identified thermophilic and non-thermophilic serine proteases with an accuracy of 95.71%. Knowledge of thermostability along with amino acid positional shifts can be vital for downstream protein engineering techniques. Thus, to emphasize the real time application of the enzyme specific classification model, a web platform has been designed. Combining the sequence data and the classification model, this prototype can allow users to align their query serine protease sequence against the custom database and identify its thermophilic nature.
Brain-computer interface (BCI) systems are usually designed specifically for each subject based on motor imagery. Therefore, the usability of these networks has become a significant challenge. The network has to be designed separately for each user, which is time-consuming for the user. Therefore, this study proposes a method by which the calibration time is significantly reduced while the classification accuracy is increased. In this method, we calibrated the features extracted from the motor imagery task by dividing the features extracted from the resting-state into both open-eye and closed-eye modes and the state in which the subject moves his eyes. The best classification accuracy was obtained using the SVM classifier using the resting-state signal in the open eye, which increased by 3.64% to 74.04%. In this paper, we also investigated the effect of recording time of the resting-state signal and the impact of eye state on the classification accuracy.
We study the strategic simplicity of stable matching mechanisms where one side has fixed preferences, termed priorities. Specifically, we ask which priorities are such that the strategyproofness of deferred acceptance (DA) can be recognized by agents unable to perform contingency reasoning, that is, \emph{when is DA obviously strategyproof} (Li, 2017)? We answer this question by completely characterizing those priorities which make DA obviously strategyproof (OSP). This solves an open problem of Ashlagi and Gonczarowski, 2018. We find that when DA is OSP, priorities are either acyclic (Ergin, 2002), a restrictive condition which allows priorities to only differ on only two agents at a time, or contain an extremely limited cyclic pattern where all priority lists are identical except for exactly two. We conclude that, for stable matching mechanisms, the tension between understandability (in the sense of OSP) and expressiveness of priorities is very high.
Mutual learning, in which multiple networks learn by sharing their knowledge, improves the performance of each network. However, the performance of ensembles of networks that have undergone mutual learning does not improve significantly from that of normal ensembles without mutual learning, even though the performance of each network has improved significantly. This may be due to the relationship between the knowledge in mutual learning and the individuality of the networks in the ensemble. In this study, we propose an ensemble method using knowledge transfer to improve the accuracy of ensembles by introducing a loss design that promotes diversity among networks in mutual learning. We use an attention map as knowledge, which represents the probability distribution and information in the middle layer of a network. There are many ways to combine networks and loss designs for knowledge transfer methods. Therefore, we use the automatic optimization of knowledge-transfer graphs to consider a variety of knowledge-transfer methods by graphically representing conventional mutual-learning and distillation methods and optimizing each element through hyperparameter search. The proposed method consists of a mechanism for constructing an ensemble in a knowledge-transfer graph, attention loss, and a loss design that promotes diversity among networks. We explore optimal ensemble learning by optimizing a knowledge-transfer graph to maximize ensemble accuracy. From exploration of graphs and evaluation experiments using the datasets of Stanford Dogs, Stanford Cars, and CUB-200-2011, we confirm that the proposed method is more accurate than a conventional ensemble method.
Many classification models produce a probability distribution as the outcome of a prediction. This information is generally compressed down to the single class with the highest associated probability. In this paper, we argue that part of the information that is discarded in this process can be in fact used to further evaluate the goodness of models, and in particular the confidence with which each prediction is made. As an application of the ideas presented in this paper, we provide a theoretical explanation of a confidence degradation phenomenon observed in the complement approach to the (Bernoulli) Naive Bayes generative model.
Online information systems currently heavily rely on the username and password traditional method for protecting information and controlling access. With the advancement in biometric technology and popularity of fields like AI and Machine Learning, biometric security is becoming increasingly popular because of the usability advantage. This paper reports how machine learning based face recognition can be integrated into a web-based system as a method of authentication to reap the benefits of improved usability. This paper includes a comparison of combinations of detection and classification algorithms with FaceNet for face recognition. The results show that a combination of MTCNN for detection, Facenet for generating embeddings, and LinearSVC for classification outperforms other combinations with a 95% accuracy. The resulting classifier is integrated into the web-based system and used for authenticating users.
The attention mechanisms have been employed in Convolutional Neural Network (CNN) to enhance the feature representation. However, existing attention mechanisms only concentrate on refining the features inside each sample and neglect the discrimination between different samples. In this paper, we propose a batch aware attention module (BA2M) for feature enrichment from a distinctive perspective. More specifically, we first get the sample-wise attention representation (SAR) by fusing the channel, local spatial and global spatial attention maps within each sample. Then, we feed the SARs of the whole batch to a normalization function to get the weights for each sample. The weights serve to distinguish the features' importance between samples in a training batch with different complexity of content. The BA2M could be embedded into different parts of CNN and optimized with the network in an end-to-end manner. The design of BA2M is lightweight with few extra parameters and calculations. We validate BA2M through extensive experiments on CIFAR-100 and ImageNet-1K for the image recognition task. The results show that BA2M can boost the performance of various network architectures and outperforms many classical attention methods. Besides, BA2M exceeds traditional methods of re-weighting samples based on the loss value.
The datasets most widely used for abusive language detection contain lists of messages, usually tweets, that have been manually judged as abusive or not by one or more annotators, with the annotation performed at message level. In this paper, we investigate what happens when the hateful content of a message is judged also based on the context, given that messages are often ambiguous and need to be interpreted in the context of occurrence. We first re-annotate part of a widely used dataset for abusive language detection in English in two conditions, i.e. with and without context. Then, we compare the performance of three classification algorithms obtained on these two types of dataset, arguing that a context-aware classification is more challenging but also more similar to a real application scenario.
Video classification and analysis is always a popular and challenging field in computer vision. It is more than just simple image classification due to the correlation with respect to the semantic contents of subsequent frames brings difficulties for video analysis. In this literature review, we summarized some state-of-the-art methods for multi-label video classification. Our goal is first to experimentally research the current widely used architectures, and then to develop a method to deal with the sequential data of frames and perform multi-label classification based on automatic content detection of video.
Remote sensing scene classification deals with the problem of classifying land use/cover of a region from images. To predict the development and socioeconomic structures of cities, the status of land use in regions is tracked by the national mapping agencies of countries. Many of these agencies use land-use types that are arranged in multiple levels. In this paper, we examined the efficiency of a hierarchically designed Convolutional Neural Network (CNN) based framework that is suitable for such arrangements. We use the NWPU-RESISC45 dataset for our experiments and arranged this data set in a two-level nested hierarchy. Each node in the designed hierarchy is trained using DenseNet-121 architectures. We provide detailed empirical analysis to compare the performances of this hierarchical scheme and its non-hierarchical counterpart, together with the individual model performances. We also evaluated the performance of the hierarchical structure statistically to validate the presented empirical results. The results of our experiments show that although individual classifiers for different sub-categories in the hierarchical scheme perform considerably well, the accumulation of the classification errors in the cascaded structure prevents its classification performance from exceeding that of the non-hierarchical deep model
Recent work has argued that classification losses utilizing softmax cross-entropy are superior not only for fixed-set classification tasks, but also by outperforming losses developed specifically for open-set tasks including few-shot learning and retrieval. Softmax classifiers have been studied using different embedding geometries -- Euclidean, hyperbolic, and spherical -- and claims have been made about the superiority of one or another, but they have not been systematically compared with careful controls. We conduct an empirical investigation of embedding geometry on softmax losses for a variety of fixed-set classification and image retrieval tasks. An interesting property observed for the spherical losses lead us to propose a probabilistic classifier based on the von Mises-Fisher distribution, and we show that it is competitive with state-of-the-art methods while producing improved out-of-the-box calibration. We provide guidance regarding the trade-offs between losses and how to choose among them.
Graph classification is a challenging research problem in many applications across a broad range of domains. In these applications, it is very common that class distribution is imbalanced. Recently, Graph Neural Network (GNN) models have achieved superior performance on various real-world datasets. Despite their success, most of current GNN models largely overlook the important setting of imbalanced class distribution, which typically results in prediction bias towards majority classes. To alleviate the prediction bias, we propose to leverage semantic structure of dataset based on the distribution of node embedding. Specifically, we present GraphDIVE, a general framework leveraging mixture of diverse experts (i.e., graph classifiers) for imbalanced graph classification. With a divide-and-conquer principle, GraphDIVE employs a gating network to partition an imbalanced graph dataset into several subsets. Then each expert network is trained based on its corresponding subset. Experiments on real-world imbalanced graph datasets demonstrate the effectiveness of GraphDIVE.
In video data, busy motion details from moving regions are conveyed within a specific frequency bandwidth in the frequency domain. Meanwhile, the rest of the frequencies of video data are encoded with quiet information with substantial redundancy, which causes low processing efficiency in existing video models that take as input raw RGB frames. In this paper, we consider allocating intenser computation for the processing of the important busy information and less computation for that of the quiet information. We design a trainable Motion Band-Pass Module (MBPM) for separating busy information from quiet information in raw video data. By embedding the MBPM into a two-pathway CNN architecture, we define a Busy-Quiet Net (BQN). The efficiency of BQN is determined by avoiding redundancy in the feature space processed by the two pathways: one operating on Quiet features of low-resolution, while the other processes Busy features. The proposed BQN outperforms many recent video processing models on Something-Something V1, Kinetics400, UCF101 and HMDB51 datasets.
The first step toward Seed Phenotyping i.e. the comprehensive assessment of complex seed traits such as growth, development, tolerance, resistance, ecology, yield, and the measurement of pa-rameters that form more complex traits is the identification of seed type. Generally, a plant re-searcher inspects the visual attributes of a seed such as size, shape, area, color and texture to identify the seed type, a process that is tedious and labor-intensive. Advances in the areas of computer vision and deep learning have led to the development of convolutional neural networks (CNN) that aid in classification using images. While they classify efficiently, a key bottleneck is the need for an extensive amount of labelled data to train the CNN before it can be put to the task of classification. The work leverages the concepts of Contrastive Learning and Domain Randomi-zation in order to achieve the same. Briefly, domain randomization is the technique of applying models trained on images containing simulated objects to real-world objects. The use of synthetic images generated from a representational sample crop of real-world images alleviates the need for a large volume of test subjects. As part of the work, synthetic image datasets of five different types of seed images namely, canola, rough rice, sorghum, soy and wheat are applied to three different self-supervised learning frameworks namely, SimCLR, Momentum Contrast (MoCo) and Build Your Own Latent (BYOL) where ResNet-50 is used as the backbone in each of the networks. When the self-supervised models are fine-tuned with only 5% of the labels from the synthetic dataset, results show that MoCo, the model that yields the best performance of the self-supervised learning frameworks in question, achieves an accuracy of 77% on the test dataset which is only ~13% less than the accuracy of 90% achieved by ResNet-50 trained on 100% of the labels.
We present pure-transformer based models for video classification, drawing upon the recent success of such models in image classification. Our model extracts spatio-temporal tokens from the input video, which are then encoded by a series of transformer layers. In order to handle the long sequences of tokens encountered in video, we propose several, efficient variants of our model which factorise the spatial- and temporal-dimensions of the input. Although transformer-based models are known to only be effective when large training datasets are available, we show how we can effectively regularise the model during training and leverage pretrained image models to be able to train on comparatively small datasets. We conduct thorough ablation studies, and achieve state-of-the-art results on multiple video classification benchmarks including Kinetics 400 and 600, Epic Kitchens, Something-Something v2 and Moments in Time, outperforming prior methods based on deep 3D convolutional networks. To facilitate further research, we release code at https://github.com/google-research/scenic/tree/main/scenic/projects/vivit
The recently developed vision transformer (ViT) has achieved promising results on image classification compared to convolutional neural networks. Inspired by this, in this paper, we study how to learn multi-scale feature representations in transformer models for image classification. To this end, we propose a dual-branch transformer to combine image patches (i.e., tokens in a transformer) of different sizes to produce stronger image features. Our approach processes small-patch and large-patch tokens with two separate branches of different computational complexity and these tokens are then fused purely by attention multiple times to complement each other. Furthermore, to reduce computation, we develop a simple yet effective token fusion module based on cross attention, which uses a single token for each branch as a query to exchange information with other branches. Our proposed cross-attention only requires linear time for both computational and memory complexity instead of quadratic time otherwise. Extensive experiments demonstrate that our approach performs better than or on par with several concurrent works on vision transformer, in addition to efficient CNN models. For example, on the ImageNet1K dataset, with some architectural changes, our approach outperforms the recent DeiT by a large margin of 2\% with a small to moderate increase in FLOPs and model parameters. Our source codes and models are available at \url{https://github.com/IBM/CrossViT}.
Artificial intelligence (AI) and specifically machine learning is making inroads into number of fields. Machine learning is replacing and/or complementing humans in a certain type of domain to make systems perform tasks more efficiently and independently. Healthcare is a worthy domain to merge with AI and Machine learning to get things to work smoother and efficiently. The X-ray based detection and classification of diseases related to chest is much needed in this modern era due to the low number of quality radiologists. This thesis focuses on the classification of Pneumonia and Tuberculosis two major chest diseases from the chest X-rays. This system provides an opinion to the user whether one is having a disease or not, thereby helping doctors and medical staff to make a quick and informed decision about the presence of disease. As compared to previous work our model can detect two types of abnormality. Our model can detect whether X-ray is normal or having abnormality which can be pneumonia and tuberculosis 92.97% accurately.