Nowadays deep learning-based methods have achieved a remarkable progress at the image classification task among a wide range of commonly used datasets (ImageNet, CIFAR, SVHN, Caltech 101, SUN397, etc.). SOTA performance on each of the mentioned datasets is obtained by careful tuning of the model architecture and training tricks according to the properties of the target data. Although this approach allows setting academic records, it is unrealistic that an average data scientist would have enough resources to build a sophisticated training pipeline for every image classification task he meets in practice. This work is focusing on reviewing the latest augmentation and regularization methods for the image classification and exploring ways to automatically choose some of the most important hyperparameters: total number of epochs, initial learning rate value and it's schedule. Having a training procedure equipped with a lightweight modern CNN architecture (like bileNetV3 or EfficientNet), sufficient level of regularization and adaptive to data learning rate schedule, we can achieve a reasonable performance on a variety of downstream image classification tasks without manual tuning of parameters to each particular task. Resulting models are computationally efficient and can be deployed to CPU using the OpenVINO toolkit. Source code is available as a part of the OpenVINO Training Extensions (https://github.com/openvinotoolkit/training_extensions).
While there has been a number of studies on Zero-Shot Learning (ZSL) for 2D images, its application to 3D data is still recent and scarce, with just a few methods limited to classification. We present the first generative approach for both ZSL and Generalized ZSL (GZSL) on 3D data, that can handle both classification and, for the first time, semantic segmentation. We show that it reaches or outperforms the state of the art on ModelNet40 classification for both inductive ZSL and inductive GZSL. For semantic segmentation, we created three benchmarks for evaluating this new ZSL task, using S3DIS, ScanNet and SemanticKITTI. Our experiments show that our method outperforms strong baselines, which we additionally propose for this task.
We present our approach to improve room classification task on floor plan maps of buildings by representing floor plans as undirected graphs and leveraging graph neural networks to predict the room categories. Rooms in the floor plans are represented as nodes in the graph with edges representing their adjacency in the map. We experiment with House-GAN dataset that consists of floor plan maps in vector format and train multilayer perceptron and graph neural networks. Our results show that graph neural networks, specifically GraphSAGE and Topology Adaptive GCN were able to achieve accuracy of 80% and 81% respectively outperforming baseline multilayer perceptron by more than 15% margin.
Deep learning has celebrated resounding successes in many application areas of relevance to the Internet of Things (IoT), such as computer vision and machine listening. These technologies must ultimately be brought directly to the edge to fully harness the power of deep learning for the IoT. The obvious challenge is that deep learning techniques can only be implemented on strictly resource-constrained edge devices if the models are radically downsized. This task relies on different model compression techniques, such as network pruning, quantization, and the recent advancement of XNOR-Net. This study examines the suitability of these techniques for audio classification on microcontrollers. We present an application of XNOR-Net for end-to-end raw audio classification and a comprehensive empirical study comparing this approach with pruning-and-quantization methods. We show that raw audio classification with XNOR yields comparable performance to regular full precision networks for small numbers of classes while reducing memory requirements 32-fold and computation requirements 58-fold. However, as the number of classes increases significantly, performance degrades, and pruning-and-quantization based compression techniques take over as the preferred technique being able to satisfy the same space constraints but requiring approximately 8x more computation. We show that these insights are consistent between raw audio classification and image classification using standard benchmark sets. To the best of our knowledge, this is the first study to apply XNOR to end-to-end audio classification and evaluate it in the context of alternative techniques. All codes are publicly available on GitHub.
Plankton are effective indicators of environmental change and ecosystem health in freshwater habitats, but collection of plankton data using manual microscopic methods is extremely labor-intensive and expensive. Automated plankton imaging offers a promising way forward to monitor plankton communities with high frequency and accuracy in real-time. Yet, manual annotation of millions of images proposes a serious challenge to taxonomists. Deep learning classifiers have been successfully applied in various fields and provided encouraging results when used to categorize marine plankton images. Here, we present a set of deep learning models developed for the identification of lake plankton, and study several strategies to obtain optimal performances,which lead to operational prescriptions for users. To this aim, we annotated into 35 classes over 17900 images of zooplankton and large phytoplankton colonies, detected in Lake Greifensee (Switzerland) with the Dual Scripps Plankton Camera. Our best models were based on transfer learning and ensembling, which classified plankton images with 98% accuracy and 93% F1 score. When tested on freely available plankton datasets produced by other automated imaging tools (ZooScan, FlowCytobot and ISIIS), our models performed better than previously used models. Our annotated data, code and classification models are freely available online.
Two-stage methods have dominated Human-Object Interaction (HOI) detection for several years. Recently, one-stage HOI detection methods have become popular. In this paper, we aim to explore the essential pros and cons of two-stage and one-stage methods. With this as the goal, we find that conventional two-stage methods mainly suffer from positioning positive interactive human-object pairs, while one-stage methods are challenging to make an appropriate trade-off on multi-task learning, i.e., object detection, and interaction classification. Therefore, a core problem is how to take the essence and discard the dregs from the conventional two types of methods. To this end, we propose a novel one-stage framework with disentangling human-object detection and interaction classification in a cascade manner. In detail, we first design a human-object pair generator based on a state-of-the-art one-stage HOI detector by removing the interaction classification module or head and then design a relatively isolated interaction classifier to classify each human-object pair. Two cascade decoders in our proposed framework can focus on one specific task, detection or interaction classification. In terms of the specific implementation, we adopt a transformer-based HOI detector as our base model. The newly introduced disentangling paradigm outperforms existing methods by a large margin, with a significant relative mAP gain of 9.32% on HICO-Det. The source codes are available at https://github.com/YueLiao/CDN.
The Influenza virus can be considered as one of the most severe viruses that can infect multiple species with often fatal consequences to the hosts. The Hemagglutinin (HA) gene of the virus can be a target for antiviral drug development realised through accurate identification of its sub-types and possible the targeted hosts. This paper focuses on accurately predicting if an Influenza type A virus can infect specific hosts, and more specifically, Human, Avian and Swine hosts, using only the protein sequence of the HA gene. In more detail, we propose encoding the protein sequences into numerical signals using the Hydrophobicity Index and subsequently utilising a Convolutional Neural Network-based predictive model. The Influenza HA protein sequences used in the proposed work are obtained from the Influenza Research Database (IRD). Specifically, complete and unique HA protein sequences were used for avian, human and swine hosts. The data obtained for this work was 17999 human-host proteins, 17667 avian-host proteins and 9278 swine-host proteins. Given this set of collected proteins, the proposed method yields as much as 10% higher accuracy for an individual class (namely, Avian) and 5% higher overall accuracy than in an earlier study. It is also observed that the accuracy for each class in this work is more balanced than what was presented in this earlier study. As the results show, the proposed model can distinguish HA protein sequences with high accuracy whenever the virus under investigation can infect Human, Avian or Swine hosts.
Poisoning attacks are a category of adversarial machine learning threats in which an adversary attempts to subvert the outcome of the machine learning systems by injecting crafted data into training data set, thus increasing the machine learning model's test error. The adversary can tamper with the data feature space, data labels, or both, each leading to a different attack strategy with different strengths. Various detection approaches have recently emerged, each focusing on one attack strategy. The Achilles heel of many of these detection approaches is their dependence on having access to a clean, untampered data set. In this paper, we propose CAE, a Classification Auto-Encoder based detector against diverse poisoned data. CAE can detect all forms of poisoning attacks using a combination of reconstruction and classification errors without having any prior knowledge of the attack strategy. We show that an enhanced version of CAE (called CAE+) does not have to employ a clean data set to train the defense model. Our experimental results on three real datasets MNIST, Fashion-MNIST and CIFAR demonstrate that our proposed method can maintain its functionality under up to 30% contaminated data and help the defended SVM classifier to regain its best accuracy.
Few-shot learning (FSL) attempts to learn with limited data. In this work, we perform the feature extraction in the Euclidean space and the geodesic distance metric on the Oblique Manifold (OM). Specially, for better feature extraction, we propose a non-parametric Region Self-attention with Spatial Pyramid Pooling (RSSPP), which realizes a trade-off between the generalization and the discriminative ability of the single image feature. Then, we embed the feature to OM as a point. Furthermore, we design an Oblique Distance-based Classifier (ODC) that achieves classification in the tangent spaces which better approximate OM locally by learnable tangency points. Finally, we introduce a new method for parameters initialization and a novel loss function in the transductive settings. Extensive experiments demonstrate the effectiveness of our algorithm and it outperforms state-of-the-art methods on the popular benchmarks: mini-ImageNet, tiered-ImageNet, and Caltech-UCSD Birds-200-2011 (CUB).
Class imbalance in a dataset is a major problem for classifiers that results in poor prediction with a high true positive rate (TPR) but a low true negative rate (TNR) for a majority positive training dataset. Generally, the pre-processing technique of oversampling of minority class(es) are used to overcome this deficiency. Our focus is on using the hybridization of Generative Adversarial Network (GAN) and Synthetic Minority Over-Sampling Technique (SMOTE) to address class imbalanced problems. We propose a novel two-phase oversampling approach involving knowledge transfer that has the synergy of SMOTE and GAN. The unrealistic or overgeneralized samples of SMOTE are transformed into realistic distribution of data by GAN where there is not enough minority class data available for GAN to process them by itself effectively. We named it SMOTified-GAN as GAN works on pre-sampled minority data produced by SMOTE rather than randomly generating the samples itself. The experimental results prove the sample quality of minority class(es) has been improved in a variety of tested benchmark datasets. Its performance is improved by up to 9\% from the next best algorithm tested on F1-score measurements. Its time complexity is also reasonable which is around $O(N^2d^2T)$ for a sequential algorithm.
We introduce a noisy channel approach for language model prompting in few-shot text classification. Instead of computing the likelihood of the label given the input (referred as direct models), channel models compute the conditional probability of the input given the label, and are thereby required to explain every word in the input. We use channel models for recently proposed few-shot learning methods with no or very limited updates to the language model parameters, via either in-context demonstration or prompt tuning. Our experiments show that, for both methods, channel models significantly outperform their direct counterparts, which we attribute to their stability, i.e., lower variance and higher worst-case accuracy. We also present extensive ablations that provide recommendations for when to use channel prompt tuning instead of other competitive methods (e.g., direct head tuning): channel prompt tuning is preferred when the number of training examples is small, labels in the training data are imbalanced, or generalization to unseen labels is required.
End-to-end intent classification using speech has numerous advantages compared to the conventional pipeline approach using automatic speech recognition (ASR), followed by natural language processing modules. It attempts to predict intent from speech without using an intermediate ASR module. However, such end-to-end framework suffers from the unavailability of large speech resources with higher acoustic variation in spoken language understanding. In this work, we exploit the scope of the transformer distillation method that is specifically designed for knowledge distillation from a transformer based language model to a transformer based speech model. In this regard, we leverage the reliable and widely used bidirectional encoder representations from transformers (BERT) model as a language model and transfer the knowledge to build an acoustic model for intent classification using the speech. In particular, a multilevel transformer based teacher-student model is designed, and knowledge distillation is performed across attention and hidden sub-layers of different transformer layers of the student and teacher models. We achieve an intent classification accuracy of 99.10% and 88.79% for Fluent speech corpus and ATIS database, respectively. Further, the proposed method demonstrates better performance and robustness in acoustically degraded condition compared to the baseline method.
Transformer achieves remarkable successes in understanding 1 and 2-dimensional signals (e.g., NLP and Image Content Understanding). As a potential alternative to convolutional neural networks, it shares merits of strong interpretability, high discriminative power on hyper-scale data, and flexibility in processing varying length inputs. However, its encoders naturally contain computational intensive operations such as pair-wise self-attention, incurring heavy computational burden when being applied on the complex 3-dimensional video signals. This paper presents Token Shift Module (i.e., TokShift), a novel, zero-parameter, zero-FLOPs operator, for modeling temporal relations within each transformer encoder. Specifically, the TokShift barely temporally shifts partial [Class] token features back-and-forth across adjacent frames. Then, we densely plug the module into each encoder of a plain 2D vision transformer for learning 3D video representation. It is worth noticing that our TokShift transformer is a pure convolutional-free video transformer pilot with computational efficiency for video understanding. Experiments on standard benchmarks verify its robustness, effectiveness, and efficiency. Particularly, with input clips of 8/12 frames, the TokShift transformer achieves SOTA precision: 79.83%/80.40% on the Kinetics-400, 66.56% on EGTEA-Gaze+, and 96.80% on UCF-101 datasets, comparable or better than existing SOTA convolutional counterparts. Our code is open-sourced in: https://github.com/VideoNetworks/TokShift-Transformer.
Tuning pre-trained language models (PLMs) with task-specific prompts has been a promising approach for text classification. Particularly, previous studies suggest that prompt-tuning has remarkable superiority in the low-data scenario over the generic fine-tuning methods with extra classifiers. The core idea of prompt-tuning is to insert text pieces, i.e., template, to the input and transform a classification problem into a masked language modeling problem, where a crucial step is to construct a projection, i.e., verbalizer, between a label space and a label word space. A verbalizer is usually handcrafted or searched by gradient descent, which may lack coverage and bring considerable bias and high variances to the results. In this work, we focus on incorporating external knowledge into the verbalizer, forming a knowledgeable prompt-tuning (KPT), to improve and stabilize prompt-tuning. Specifically, we expand the label word space of the verbalizer using external knowledge bases (KBs) and refine the expanded label word space with the PLM itself before predicting with the expanded label word space. Extensive experiments on zero and few-shot text classification tasks demonstrate the effectiveness of knowledgeable prompt-tuning.
In the last couple of decades, tremendous progress has been achieved in developing robotic telescopes and, as a result, sky surveys (both terrestrial and space) have become the source of a substantial amount of new observational data. These data contain a lot of information about binary stars, hidden in their light curves. With the huge amount of astronomical data gathered, it is not reasonable to expect all the data to be manually processed and analyzed. Therefore, in this paper, we focus on the automatic classification of eclipsing binary stars using deep learning methods. Our classifier provides a tool for the categorization of light curves of binary stars into two classes: detached and over-contact. We used the ELISa software to obtain synthetic data, which we then used for the training of the classifier. For evaluation purposes, we collected 100 light curves of observed binary stars, in order to evaluate a number of classifiers. We evaluated semi-detached eclipsing binary stars as detached. The best-performing classifier combines bidirectional Long Short-Term Memory (LSTM) and a one-dimensional convolutional neural network, which achieved 98% accuracy on the evaluation set. Omitting semi-detached eclipsing binary stars, we could obtain 100% accuracy in classification.
Deep extreme classification (XC) seeks to train deep architectures that can tag a data point with its most relevant subset of labels from an extremely large label set. The core utility of XC comes from predicting labels that are rarely seen during training. Such rare labels hold the key to personalized recommendations that can delight and surprise a user. However, the large number of rare labels and small amount of training data per rare label offer significant statistical and computational challenges. State-of-the-art deep XC methods attempt to remedy this by incorporating textual descriptions of labels but do not adequately address the problem. This paper presents ECLARE, a scalable deep learning architecture that incorporates not only label text, but also label correlations, to offer accurate real-time predictions within a few milliseconds. Core contributions of ECLARE include a frugal architecture and scalable techniques to train deep models along with label correlation graphs at the scale of millions of labels. In particular, ECLARE offers predictions that are 2 to 14% more accurate on both publicly available benchmark datasets as well as proprietary datasets for a related products recommendation task sourced from the Bing search engine. Code for ECLARE is available at https://github.com/Extreme-classification/ECLARE.
Extreme multi-label classification (XML) involves tagging a data point with its most relevant subset of labels from an extremely large label set, with several applications such as product-to-product recommendation with millions of products. Although leading XML algorithms scale to millions of labels, they largely ignore label meta-data such as textual descriptions of the labels. On the other hand, classical techniques that can utilize label metadata via representation learning using deep networks struggle in extreme settings. This paper develops the DECAF algorithm that addresses these challenges by learning models enriched by label metadata that jointly learn model parameters and feature representations using deep networks and offer accurate classification at the scale of millions of labels. DECAF makes specific contributions to model architecture design, initialization, and training, enabling it to offer up to 2-6% more accurate prediction than leading extreme classifiers on publicly available benchmark product-to-product recommendation datasets, such as LF-AmazonTitles-1.3M. At the same time, DECAF was found to be up to 22x faster at inference than leading deep extreme classifiers, which makes it suitable for real-time applications that require predictions within a few milliseconds. The code for DECAF is available at the following URL https://github.com/Extreme-classification/DECAF.
Attempts of learning from hierarchical taxonomies in computer vision have been mostly focusing on image classification. Though ways of best harvesting learning improvements from hierarchies in classification are far from being solved, there is a need to target these problems in other vision tasks such as object detection. As progress on the classification side is often dependent on hierarchical cross-entropy losses, novel detection architectures using sigmoid as an output function instead of softmax cannot easily apply these advances, requiring novel methods in detection. In this work we establish a theoretical framework based on probability and set theory for extracting parent predictions and a hierarchical loss that can be used across tasks, showing results across classification and detection benchmarks and opening up the possibility of hierarchical learning for sigmoid-based detection architectures.
Fine-grained image classification involves identifying different subcategories of a class which possess very subtle discriminatory features. Fine-grained datasets usually provide bounding box annotations along with class labels to aid the process of classification. However, building large scale datasets with such annotations is a mammoth task. Moreover, this extensive annotation is time-consuming and often requires expertise, which is a huge bottleneck in building large datasets. On the other hand, self-supervised learning (SSL) exploits the freely available data to generate supervisory signals which act as labels. The features learnt by performing some pretext tasks on huge unlabelled data proves to be very helpful for multiple downstream tasks. Our idea is to leverage self-supervision such that the model learns useful representations of fine-grained image classes. We experimented with 3 kinds of models: Jigsaw solving as pretext task, adversarial learning (SRGAN) and contrastive learning based (SimCLR) model. The learned features are used for downstream tasks such as fine-grained image classification. Our code is available at http://github.com/rush2406/Self-Supervised-Learning-for-Fine-grained-Image-Classification
Heart disease is the most common reason for human mortality that causes almost one-third of deaths throughout the world. Detecting the disease early increases the chances of survival of the patient and there are several ways a sign of heart disease can be detected early. This research proposes to convert cleansed and normalized heart sound into visual mel scale spectrograms and then using visual domain transfer learning approaches to automatically extract features and categorize between heart sounds. Some of the previous studies found that the spectrogram of various types of heart sounds is visually distinguishable to human eyes, which motivated this study to experiment on visual domain classification approaches for automated heart sound classification. It will use convolution neural network-based architectures i.e. ResNet, MobileNetV2, etc as the automated feature extractors from spectrograms. These well-accepted models in the image domain showed to learn generalized feature representations of cardiac sounds collected from different environments with varying amplitude and noise levels. Model evaluation criteria used were categorical accuracy, precision, recall, and AUROC as the chosen dataset is unbalanced. The proposed approach has been implemented on datasets A and B of the PASCAL heart sound collection and resulted in ~ 90% categorical accuracy and AUROC of ~0.97 for both sets.