To facilitate implementation of high-accuracy deep neural networks especially on resource-constrained devices, maintaining low computation requirements is crucial. Using very deep models for classification purposes not only decreases the neural network training speed and increases the inference time, but also need more data for higher prediction accuracy and to mitigate false positives. In this paper, we propose an efficient and lightweight deep classification ensemble structure based on a combination of simple color features, which is particularly designed for "high-accuracy" image classifications with low false positives. We designed, implemented, and evaluated our approach for explosion detection use-case applied to images and videos. Our evaluation results based on a large test test show considerable improvements on the prediction accuracy compared to the popular ResNet-50 model, while benefiting from 7.64x faster inference and lower computation cost. While we applied our approach to explosion detection, our approach is general and can be applied to other similar classification use cases as well. Given the insight gained from our experiments, we hence propose a "think small, think many" philosophy in classification scenarios: that transforming a single, large, monolithic deep model into a verification-based step model ensemble of multiple small, simple, lightweight models with narrowed-down color spaces can possibly lead to predictions with higher accuracy.
We introduce a novel fine-grained dataset and benchmark, the Danish Fungi 2020 (DF20). The dataset, constructed from observations submitted to the Atlas of Danish Fungi, is unique in its taxonomy-accurate class labels, small number of errors, highly unbalanced long-tailed class distribution, rich observation metadata, and well-defined class hierarchy. DF20 has zero overlap with ImageNet, allowing unbiased comparison of models fine-tuned from publicly available ImageNet checkpoints. The proposed evaluation protocol enables testing the ability to improve classification using metadata -- e.g. precise geographic location, habitat, and substrate, facilitates classifier calibration testing, and finally allows to study the impact of the device settings on the classification performance. Experiments using Convolutional Neural Networks (CNN) and the recent Vision Transformers (ViT) show that DF20 presents a challenging task. Interestingly, ViT achieves results superior to CNN baselines with 80.45% accuracy and 0.743 macro F1 score, reducing the CNN error by 9% and 12% respectively. A simple procedure for including metadata into the decision process improves the classification accuracy by more than 2.95 percentage points, reducing the error rate by 15%. The source code for all methods and experiments is available at https://sites.google.com/view/danish-fungi-dataset.
This paper proposes a new distance metric function, called Z distance, for KNN classification. The Z distance function is not a geometric direct-line distance between two data points. It gives a consideration to the class attribute of a training dataset when measuring the affinity between data points. Concretely speaking, the Z distance of two data points includes their class center distance and real distance. And its shape looks like "Z". In this way, the affinity of two data points in the same class is always stronger than that in different classes. Or, the intraclass data points are always closer than those interclass data points. We evaluated the Z distance with experiments, and demonstrated that the proposed distance function achieved better performance in KNN classification.
Two of the main challenges for cropland classification by satellite time-series images are insufficient ground-truth data and inaccessibility of high-quality hyperspectral images for under-developed areas. Unlabeled medium-resolution satellite images are abundant, but how to benefit from them is an open question. We will show how to leverage their potential for cropland classification using self-supervised tasks. Self-supervision is an approach where we provide simple training signals for the samples, which are apparent from the data's structure. Hence, they are cheap to acquire and explain a simple concept about the data. We introduce three self-supervised tasks for cropland classification. They reduce epistemic uncertainty, and the resulting model shows superior accuracy in a wide range of settings compared to SVM and Random Forest. Subsequently, we use the self-supervised tasks to perform unsupervised domain adaptation and benefit from the labeled samples in other regions. It is crucial to know what information to transfer to avoid degrading the performance. We show how to automate the information selection and transfer process in cropland classification even when the source and target areas have a very different feature distribution. We improved the model by about 24% compared to a baseline architecture without any labeled sample in the target domain. Our method is amenable to gradual improvement, works with medium-resolution satellite images, and does not require complicated models. Code and data are available.
The problem of classifying high-dimensional shapes in real-world data grows in complexity as the dimension of the space increases. For the case of identifying convex shapes of different geometries, a new classification framework has recently been proposed in which the intersections of a set of one-dimensional representations, called rays, with the boundaries of the shape are used to identify the specific geometry. This ray-based classification (RBC) has been empirically verified using a synthetic dataset of two- and three-dimensional shapes (Zwolak et al. in Proceedings of Third Workshop on Machine Learning and the Physical Sciences (NeurIPS 2020), Vancouver, Canada [December 11, 2020], arXiv:2010.00500, 2020) and, more recently, has also been validated experimentally (Zwolak et al., PRX Quantum 2:020335, 2021). Here, we establish a bound on the number of rays necessary for shape classification, defined by key angular metrics, for arbitrary convex shapes. For two dimensions, we derive a lower bound on the number of rays in terms of the shape's length, diameter, and exterior angles. For convex polytopes in $\mathbb{R}^N$, we generalize this result to a similar bound given as a function of the dihedral angle and the geometrical parameters of polygonal faces. This result enables a different approach for estimating high-dimensional shapes using substantially fewer data elements than volumetric or surface-based approaches.
Object detection is a typical multi-task learning application, which optimizes classification and regression simultaneously. However, classification loss always dominates the multi-task loss in anchor-based methods, hampering the consistent and balanced optimization of the tasks. In this paper, we find that shifting the bounding boxes can change the division of positive and negative samples in classification, meaning classification depends on regression. Moreover, we summarize three important conclusions about fine-tuning loss weights, considering different datasets, optimizers and regression loss functions. Based on the above conclusions, we propose Adaptive Loss Weight Adjustment(ALWA) to solve the imbalance in optimizing anchor-based methods according to statistical characteristics of losses. By incorporating ALWA into previous state-of-the-art detectors, we achieve a significant performance gain on PASCAL VOC and MS COCO, even with L1, SmoothL1 and CIoU loss. The code is available at https://github.com/ywx-hub/ALWA.
Hyperspectral images (HSI) consist of rich spatial and spectral information, which can potentially be used for several applications. However, noise, band correlations and high dimensionality restrict the applicability of such data. This is recently addressed using creative deep learning network architectures such as ResNet, SSRN, and A2S2K. However, the last layer, i.e the classification layer, remains unchanged and is taken to be the softmax classifier. In this article, we propose to use a watershed classifier. Watershed classifier extends the watershed operator from Mathematical Morphology for classification. In its vanilla form, the watershed classifier does not have any trainable parameters. In this article, we propose a novel approach to train deep learning networks to obtain representations suitable for the watershed classifier. The watershed classifier exploits the connectivity patterns, a characteristic of HSI datasets, for better inference. We show that exploiting such characteristics allows the Triplet-Watershed to achieve state-of-art results in supervised and semi-supervised contexts. These results are validated on Indianpines (IP), University of Pavia (UP), Kennedy Space Center (KSC) and University of Houston (UH) datasets, relying on simple convnet architecture using a quarter of parameters compared to previous state-of-the-art networks. The source code for reproducing the experiments and supplementary material (high resolution images) is available at https://github.com/ac20/TripletWatershed Code.
This work empirically demonstrates the ability of Text Graph Convolutional Network (Text GCN) to outperform traditional natural language processing benchmarks for the task of semi-supervised Swahili news classification. In particular, we focus our experimentation on the sparsely-labelled semi-supervised context which is representative of the practical constraints facing low-resourced African languages. We follow up on this result by introducing a variant of the Text GCN model which utilises a bag of words embedding rather than a naive one-hot encoding to reduce the memory footprint of Text GCN whilst demonstrating similar predictive performance.
Node classification is an important research topic in graph learning. Graph neural networks (GNNs) have achieved state-of-the-art performance of node classification. However, existing GNNs address the problem where node samples for different classes are balanced; while for many real-world scenarios, some classes may have much fewer instances than others. Directly training a GNN classifier in this case would under-represent samples from those minority classes and result in sub-optimal performance. Therefore, it is very important to develop GNNs for imbalanced node classification. However, the work on this is rather limited. Hence, we seek to extend previous imbalanced learning techniques for i.i.d data to the imbalanced node classification task to facilitate GNN classifiers. In particular, we choose to adopt synthetic minority over-sampling algorithms, as they are found to be the most effective and stable. This task is non-trivial, as previous synthetic minority over-sampling algorithms fail to provide relation information for newly synthesized samples, which is vital for learning on graphs. Moreover, node attributes are high-dimensional. Directly over-sampling in the original input domain could generates out-of-domain samples, which may impair the accuracy of the classifier. We propose a novel framework, GraphSMOTE, in which an embedding space is constructed to encode the similarity among the nodes. New samples are synthesize in this space to assure genuineness. In addition, an edge generator is trained simultaneously to model the relation information, and provide it for those new samples. This framework is general and can be easily extended into different variations. The proposed framework is evaluated using three different datasets, and it outperforms all baselines with a large margin.
Band selection refers to the process of choosing the most relevant bands in a hyperspectral image. By selecting a limited number of optimal bands, we aim at speeding up model training, improving accuracy, or both. It reduces redundancy among spectral bands while trying to preserve the original information of the image. By now many efforts have been made to develop unsupervised band selection approaches, of which the majority are heuristic algorithms devised by trial and error. In this paper, we are interested in training an intelligent agent that, given a hyperspectral image, is capable of automatically learning policy to select an optimal band subset without any hand-engineered reasoning. To this end, we frame the problem of unsupervised band selection as a Markov decision process, propose an effective method to parameterize it, and finally solve the problem by deep reinforcement learning. Once the agent is trained, it learns a band-selection policy that guides the agent to sequentially select bands by fully exploiting the hyperspectral image and previously picked bands. Furthermore, we propose two different reward schemes for the environment simulation of deep reinforcement learning and compare them in experiments. This, to the best of our knowledge, is the first study that explores a deep reinforcement learning model for hyperspectral image analysis, thus opening a new door for future research and showcasing the great potential of deep reinforcement learning in remote sensing applications. Extensive experiments are carried out on four hyperspectral data sets, and experimental results demonstrate the effectiveness of the proposed method.
Retinal artery/vein (A/V) classification is a critical technique for diagnosing diabetes and cardiovascular diseases. Although deep learning based methods achieve impressive results in A/V classification, their performances usually degrade severely when being directly applied to another database, due to the domain shift, e.g., caused by the variations in imaging protocols. In this paper, we propose a novel vessel-mixing based consistency regularization framework, for cross-domain learning in retinal A/V classification. Specially, to alleviate the severe bias to source domain, based on the label smooth prior, the model is regularized to give consistent predictions for unlabeled target-domain inputs that are under perturbation. This consistency regularization implicitly introduces a mechanism where the model and the perturbation is opponent to each other, where the model is pushed to be robust enough to cope with the perturbation. Thus, we investigate a more difficult opponent to further inspire the robustness of model, in the scenario of retinal A/V, called vessel-mixing perturbation. Specially, it effectively disturbs the fundus images especially the vessel structures by mixing two images regionally. We conduct extensive experiments on cross-domain A/V classification using four public datasets, which are collected by diverse institutions and imaging devices. The results demonstrate that our method achieves the state-of-the-art cross-domain performance, which is also close to the upper bound obtained by fully supervised learning on target domain.
Object detection involves two sub-tasks, i.e. localizing objects in an image and classifying them into various categories. For existing CNN-based detectors, we notice the widespread divergence between localization and classification, which leads to degradation in performance. In this work, we propose a mutual learning framework to modulate the two tasks. In particular, the two tasks are forced to learn from each other with a novel mutual labeling strategy. Besides, we introduce a simple yet effective IoU rescoring scheme, which further reduces the divergence. Moreover, we define a Spearman rank correlation-based metric to quantify the divergence, which correlates well with the detection performance. The proposed approach is general-purpose and can be easily injected into existing detectors such as FCOS and RetinaNet. We achieve a significant performance gain over the baseline detectors on the COCO dataset.
When fine-tuning pretrained models for classification, researchers either use a generic model head or a task-specific prompt for prediction. Proponents of prompting have argued that prompts provide a method for injecting task-specific guidance, which is beneficial in low-data regimes. We aim to quantify this benefit through rigorous testing of prompts in a fair setting: comparing prompted and head-based fine-tuning in equal conditions across many tasks and data sizes. By controlling for many sources of advantage, we find that prompting does indeed provide a benefit, and that this benefit can be quantified per task. Results show that prompting is often worth 100s of data points on average across classification tasks.
Deep learning models with large learning capacities often overfit to medical imaging datasets. This is because training sets are often relatively small due to the significant time and financial costs incurred in medical data acquisition and labelling. Data augmentation is therefore often used to expand the availability of training data and to increase generalization. However, augmentation strategies are often chosen on an ad-hoc basis without justification. In this paper, we present an augmentation policy search method with the goal of improving model classification performance. We include in the augmentation policy search additional transformations that are often used in medical image analysis and evaluate their performance. In addition, we extend the augmentation policy search to include non-linear mixed-example data augmentation strategies. Using these learned policies, we show that principled data augmentation for medical image model training can lead to significant improvements in ultrasound standard plane detection, with an an average F1-score improvement of 7.0% overall over naive data augmentation strategies in ultrasound fetal standard plane classification. We find that the learned representations of ultrasound images are better clustered and defined with optimized data augmentation.
Alzheimer's disease (AD) is the most common age-related dementia. It remains a challenge to identify the individuals at risk of dementia for precise management. Brain MRI offers a noninvasive biomarker to detect brain aging. Previous evidence shows that the brain structural change detected by diffusion MRI is associated with dementia. Mounting studies has conceptualised the brain as a complex network, which has shown the utility of this approach in characterising various neurological and psychiatric disorders. Therefore, the structural connectivity shows promise in dementia classification. The proposed BrainNetGAN is a generative adversarial network variant to augment the brain structural connectivity matrices for binary dementia classification tasks. Structural connectivity matrices between separated brain regions are constructed using tractography on diffusion MRI data. The BrainNetGAN model is trained to generate fake brain connectivity matrices, which are expected to reflect latent distribution of the real brain network data. Finally, a convolutional neural network classifier is proposed for binary dementia classification. Numerical results show that the binary classification performance in the testing set was improved using the BrainNetGAN augmented dataset. The proposed methodology allows quick synthesis of an arbitrary number of augmented connectivity matrices and can be easily transferred to similar classification tasks.
This paper proposes a meta-learning approach to evolving a parametrized loss function, which is called Meta-Loss Network (MLN), for training the image classification learning on small datasets. In our approach, the MLN is embedded in the framework of classification learning as a differentiable objective function. The MLN is evolved with the Evolutionary Strategy algorithm (ES) to an optimized loss function, such that a classifier, which optimized to minimize this loss, will achieve a good generalization effect. A classifier learns on a small training dataset to minimize MLN with Stochastic Gradient Descent (SGD), and then the MLN is evolved with the precision of the small-dataset-updated classifier on a large validation dataset. In order to evaluate our approach, the MLN is trained with a large number of small sample learning tasks sampled from FashionMNIST and tested on validation tasks sampled from FashionMNIST and CIFAR10. Experiment results demonstrate that the MLN effectively improved generalization compared to classical cross-entropy error and mean squared error.
learning algorithms. In this paper, we review the classification algorithms used in the health care system (chronic diseases) and present the neural network-based Ensemble learning method. We briefly describe the commonly used algorithms and describe their critical properties. Materials and Methods: In this study, modern classification algorithms used in healthcare, examine the principles of these methods and guidelines, and to accurately diagnose and predict chronic diseases, superior machine learning algorithms with the neural network-based ensemble learning Is used. To do this, we use experimental data, real data on chronic patients (diabetes, heart, cancer) available on the UCI site. Results: We found that group algorithms designed to diagnose chronic diseases can be more effective than baseline algorithms. It also identifies several challenges to further advancing the classification of machine learning in the diagnosis of chronic diseases. Conclusion: The results show the high performance of the neural network-based Ensemble learning approach for the diagnosis and prediction of chronic diseases, which in this study reached 98.5, 99, and 100% accuracy, respectively.
Few-shot text classification is a fundamental NLP task in which a model aims to classify text into a large number of categories, given only a few training examples per category. This paper explores data augmentation -- a technique particularly suitable for training with limited data -- for this few-shot, highly-multiclass text classification setting. On four diverse text classification tasks, we find that common data augmentation techniques can improve the performance of triplet networks by up to 3.0% on average. To further boost performance, we present a simple training strategy called curriculum data augmentation, which leverages curriculum learning by first training on only original examples and then introducing augmented data as training progresses. We explore a two-stage and a gradual schedule, and find that, compared with standard single-stage training, curriculum data augmentation trains faster, improves performance, and remains robust to high amounts of noising from augmentation.
Deep Convolutional Neural Network (DCNN) and Transformer have achieved remarkable successes in image recognition. However, their performance in fine-grained image recognition is still difficult to meet the requirements of actual needs. This paper proposes a Sequence Random Network (SRN) to enhance the performance of DCNN. The output of DCNN is one-dimensional features. This one-dimensional feature abstractly represents image information, but it does not express well the detailed information of image. To address this issue, we use the proposed SRN which composed of BiLSTM and several Tanh-Dropout blocks (called BiLSTM-TDN), to further process DCNN one-dimensional features for highlighting the detail information of image. After the feature transform by BiLSTM-TDN, the recognition performance has been greatly improved. We conducted the experiments on six fine-grained image datasets. Except for FGVC-Aircraft, the accuracies of the proposed methods on the other datasets exceeded 99%. Experimental results show that BiLSTM-TDN is far superior to the existing state-of-the-art methods. In addition to DCNN, BiLSTM-TDN can also be extended to other models, such as Transformer.
This paper investigates whether the power of the models pre-trained on text data, such as BERT, can be transferred to general token sequence classification applications. To verify pre-trained models' transferability, we test the pre-trained models on text classification tasks with meanings of tokens mismatches, and real-world non-text token sequence classification data, including amino acid, DNA, and music. We find that even on non-text data, the models pre-trained on text converge faster, perform better than the randomly initialized models, and only slightly worse than the models using task-specific knowledge. We also find that the representations of the text and non-text pre-trained models share non-trivial similarities.