Annotated training data insufficiency remains to be one of the challenges of applying deep learning in medical data classification problems. Transfer learning from an already trained deep convolutional network can be used to reduce the cost of training from scratch and to train with small training data for deep learning. This raises the question of whether we can use transfer learning to overcome the training data insufficiency problem in deep learning based medical data classifications. Deep convolutional networks have been achieving high performance results on the ImageNet Large Scale Visual Recognition Competition (ILSVRC) image classification challenge. One example is the Inception-V3 model that was the first runner up on the ILSVRC 2015 challenge. Inception modules that help to extract different sized features of input images in one level of convolution are the unique features of the Inception-V3. In this work, we have used a pretrained Inception-V3 model to take advantage of its Inception modules for Diabetic Retinopathy detection. In order to tackle the labelled data insufficiency problem, we sub-sampled a smaller version of the Kaggle Diabetic Retinopathy classification challenge dataset for model training, and tested the model's accuracy on a previously unseen data subset. Our technique could be used in other deep learning based medical image classification problems facing the challenge of labeled training data insufficiency.
Deep learning forms a hierarchical network structure for representation of multiple input features. The adaptive structural learning method of Deep Belief Network (DBN) can realize a high classification capability while searching the optimal network structure during the training. The method can find the optimal number of hidden neurons for given input data in a Restricted Boltzmann Machine (RBM) by neuron generation-annihilation algorithm. Moreover, it can generate a new hidden layer in DBN by the layer generation algorithm to actualize a deep data representation. The proposed method showed higher classification accuracy for image benchmark data sets than several deep learning methods including well-known CNN methods. In this paper, a new object detection method for the DBN architecture is proposed for localization and category of objects. The method is a task for finding semantic objects in images as Bounding Box (B-Box). To investigate the effectiveness of the proposed method, the adaptive structural learning of DBN and the object detection were evaluated on the Chest X-ray image benchmark data set (CXR8), which is one of the most commonly accessible radio-logical examination for many lung diseases. The proposed method showed higher performance for both classification (more than 94.5% classification for test data) and localization (more than 90.4% detection for test data) than the other CNN methods.
Insufficient prior knowledge of a captured hyperspectral image (HSI) scene may lead the experts or the automatic labeling systems to offer incorrect labels or ambiguous labels (i.e., assigning each training sample to a group of candidate labels, among which only one of them is valid; this is also known as partial label learning) during the labeling process. Accordingly, how to learn from such data with ambiguous labels is a problem of great practical importance. In this letter, we propose a novel superpixelwise low-rank approximation (LRA)-based partial label learning method, namely SLAP, which is the first to take into account partial label learning in HSI classification. SLAP is mainly composed of two phases: disambiguating the training labels and acquiring the predictive model. Specifically, in the first phase, we propose a superpixelwise LRA-based model, preparing the affinity graph for the subsequent label propagation process while extracting the discriminative representation to enhance the following classification task of the second phase. Then to disambiguate the training labels, label propagation propagates the labeling information via the affinity graph of training pixels. In the second phase, we take advantage of the resulting disambiguated training labels and the discriminative representations to enhance the classification performance. The extensive experiments validate the advantage of the proposed SLAP method over state-of-the-art methods.
Insufficient prior knowledge of a captured hyperspectral image (HSI) scene may lead the experts or the automatic labeling systems to offer incorrect labels or ambiguous labels (i.e., assigning each training sample to a group of candidate labels, among which only one of them is valid; this is also known as partial label learning) during the labeling process. Accordingly, how to learn from such data with ambiguous labels is a problem of great practical importance. In this paper, we propose a novel superpixelwise low-rank approximation (LRA)-based partial label learning method, namely SLAP, which is the first to take into account partial label learning in HSI classification. SLAP is mainly composed of two phases: disambiguating the training labels and acquiring the predictive model. Specifically, in the first phase, we propose a superpixelwise LRA-based model, preparing the affinity graph for the subsequent label propagation process while extracting the discriminative representation to enhance the following classification task of the second phase. Then to disambiguate the training labels, label propagation propagates the labeling information via the affinity graph of training pixels. In the second phase, we take advantage of the resulting disambiguated training labels and the discriminative representations to enhance the classification performance. The extensive experiments validate the advantage of the proposed SLAP method over state-of-the-art methods.
Zero shot learning in Image Classification refers to the setting where images from some novel classes are absent in the training data but other information such as natural language descriptions or attribute vectors of the classes are available. This setting is important in the real world since one may not be able to obtain images of all the possible classes at training. While previous approaches have tried to model the relationship between the class attribute space and the image space via some kind of a transfer function in order to model the image space correspondingly to an unseen class, we take a different approach and try to generate the samples from the given attributes, using a conditional variational autoencoder, and use the generated samples for classification of the unseen classes. By extensive testing on four benchmark datasets, we show that our model outperforms the state of the art, particularly in the more realistic generalized setting, where the training classes can also appear at the test time along with the novel classes.
Visual document understanding (VDU) has rapidly advanced with the development of powerful multi-modal language models. However, these models typically require extensive document pre-training data to learn intermediate representations and often suffer a significant performance drop in real-world online industrial settings. A primary issue is their heavy reliance on OCR engines to extract local positional information within document pages, which limits the models' ability to capture global information and hinders their generalizability, flexibility, and robustness. In this paper, we introduce GlobalDoc, a cross-modal transformer-based architecture pre-trained in a self-supervised manner using three novel pretext objective tasks. GlobalDoc improves the learning of richer semantic concepts by unifying language and visual representations, resulting in more transferable models. For proper evaluation, we also propose two novel document-level downstream VDU tasks, Few-Shot Document Image Classification (DIC) and Content-based Document Image Retrieval (DIR), designed to simulate industrial scenarios more closely. Extensive experimentation has been conducted to demonstrate GlobalDoc's effectiveness in practical settings.
During a disaster event, images shared on social media helps crisis managers gain situational awareness and assess incurred damages, among other response tasks. Recent advances in computer vision and deep neural networks have enabled the development of models for real-time image classification for a number of tasks, including detecting crisis incidents, filtering irrelevant images, classifying images into specific humanitarian categories, and assessing the severity of damage. Despite several efforts, past works mainly suffer from limited resources (i.e., labeled images) available to train more robust deep learning models. In this study, we propose new datasets for disaster type detection, and informativeness classification, and damage severity assessment. Moreover, we relabel existing publicly available datasets for new tasks. We identify exact- and near-duplicates to form non-overlapping data splits, and finally consolidate them to create larger datasets. In our extensive experiments, we benchmark several state-of-the-art deep learning models and achieve promising results. We release our datasets and models publicly, aiming to provide proper baselines as well as to spur further research in the crisis informatics community.
Among applications of deep learning (DL) involving low cost sensors, remote image classification involves a physical channel that separates edge sensors and cloud classifiers. Traditional DL models must be divided between an encoder for the sensor and the decoder + classifier at the edge server. An important challenge is to effectively train such distributed models when the connecting channels have limited rate/capacity. Our goal is to optimize DL models such that the encoder latent requires low channel bandwidth while still delivers feature information for high classification accuracy. This work proposes a three-step joint learning strategy to guide encoders to extract features that are compact, discriminative, and amenable to common augmentations/transformations. We optimize latent dimension through an initial screening phase before end-to-end (E2E) training. To obtain an adjustable bit rate via a single pre-deployed encoder, we apply entropy-based quantization and/or manual truncation on the latent representations. Tests show that our proposed method achieves accuracy improvement of up to 1.5% on CIFAR-10 and 3% on CIFAR-100 over conventional E2E cross-entropy training.
Vision-language models (VLMs) offer a promising paradigm for image classification by comparing the similarity between images and class embeddings. A critical challenge lies in crafting precise textual representations for class names. While previous studies have leveraged recent advancements in large language models (LLMs) to enhance these descriptors, their outputs often suffer from ambiguity and inaccuracy. We attribute this to two primary factors: 1) the reliance on single-turn textual interactions with LLMs, leading to a mismatch between generated text and visual concepts for VLMs; 2) the oversight of the inter-class relationships, resulting in descriptors that fail to differentiate similar classes effectively. In this paper, we propose a novel framework that integrates LLMs and VLMs to find the optimal class descriptors. Our training-free approach develops an LLM-based agent with an evolutionary optimization strategy to iteratively refine class descriptors. We demonstrate our optimized descriptors are of high quality which effectively improves classification accuracy on a wide range of benchmarks. Additionally, these descriptors offer explainable and robust features, boosting performance across various backbone models and complementing fine-tuning-based methods.
During the process of classifying Hyperspectral Image (HSI), every pixel sample is categorized under a land-cover type. CNN-based techniques for HSI classification have notably advanced the field by their adept feature representation capabilities. However, acquiring deep features remains a challenge for these CNN-based methods. In contrast, transformer models are adept at extracting high-level semantic features, offering a complementary strength. This paper's main contribution is the introduction of an HSI classification model that includes two convolutional blocks, a Gate-Shift-Fuse (GSF) block and a transformer block. This model leverages the strengths of CNNs in local feature extraction and transformers in long-range context modelling. The GSF block is designed to strengthen the extraction of local and global spatial-spectral features. An effective attention mechanism module is also proposed to enhance the extraction of information from HSI cubes. The proposed method is evaluated on four well-known datasets (the Indian Pines, Pavia University, WHU-WHU-Hi-LongKou and WHU-Hi-HanChuan), demonstrating that the proposed framework achieves superior results compared to other models.
Semi-supervised learning is attracting blooming attention, due to its success in combining unlabeled data. However, pseudo-labeling-based semi-supervised approaches suffer from two problems in image classification: (1) Existing methods might fail to adopt suitable thresholds since they either use a pre-defined/fixed threshold or an ad-hoc threshold adjusting scheme, resulting in inferior performance and slow convergence. (2) Discarding unlabeled data with confidence below the thresholds results in the loss of discriminating information. To solve these issues, we develop an effective method to make sufficient use of unlabeled data. Specifically, we design a self adaptive threshold pseudo-labeling strategy, which thresholds for each class can be dynamically adjusted to increase the number of reliable samples. Meanwhile, in order to effectively utilise unlabeled data with confidence below the thresholds, we propose an unreliable sample contrastive loss to mine the discriminative information in low-confidence samples by learning the similarities and differences between sample features. We evaluate our method on several classification benchmarks under partially labeled settings and demonstrate its superiority over the other approaches.
Material classification in natural settings is a challenge due to complex interplay of geometry, reflectance properties, and illumination. Previous work on material classification relies strongly on hand-engineered features of visual samples. In this work we use a Convolutional Neural Network (convnet) that learns descriptive features for the specific task of material recognition. Specifically, transfer learning from the task of object recognition is exploited to more effectively train good features for material classification. The approach of transfer learning using convnets yields significantly higher recognition rates when compared to previous state-of-the-art approaches. We then analyze the relative contribution of reflectance and shading information by a decomposition of the image into its intrinsic components. The use of convnets for material classification was hindered by the strong demand for sufficient and diverse training data, even with transfer learning approaches. Therefore, we present a new data set containing approximately 10k images divided into 10 material categories.
Federated learning (FL) enables collaborative learning of a deep learning model without sharing the data of participating sites. FL in medical image analysis tasks is relatively new and open for enhancements. In this study, we propose FedDropoutAvg, a new federated learning approach for training a generalizable model. The proposed method takes advantage of randomness, both in client selection and also in federated averaging process. We compare FedDropoutAvg to several algorithms in an FL scenario for real-world multi-site histopathology image classification task. We show that with FedDropoutAvg, the final model can achieve performance better than other FL approaches and closer to a classical deep learning model that requires all data to be shared for centralized training. We test the trained models on a large dataset consisting of 1.2 million image tiles from 21 different centers. To evaluate the generalization ability of the proposed approach, we use held-out test sets from centers whose data was used in the FL and for unseen data from other independent centers whose data was not used in the federated training. We show that the proposed approach is more generalizable than other state-of-the-art federated training approaches. To the best of our knowledge, ours is the first study to use a randomized client and local model parameter selection procedure in a federated setting for a medical image analysis task.
Weakly supervised object localization (WSOL) is a challenging task aiming to localize objects with only image-level supervision. Recent works apply visual transformer to WSOL and achieve significant success by exploiting the long-range feature dependency in self-attention mechanism. However, existing transformer-based methods synthesize the classification feature maps as the localization map, which leads to optimization conflicts between classification and localization tasks. To address this problem, we propose to learn a task-specific spatial-aware token (SAT) to condition localization in a weakly supervised manner. Specifically, a spatial token is first introduced in the input space to aggregate representations for localization task. Then a spatial aware attention module is constructed, which allows spatial token to generate foreground probabilities of different patches by querying and to extract localization knowledge from the classification task. Besides, for the problem of sparse and unbalanced pixel-level supervision obtained from the image-level label, two spatial constraints, including batch area loss and normalization loss, are designed to compensate and enhance this supervision. Experiments show that the proposed SAT achieves state-of-the-art performance on both CUB-200 and ImageNet, with 98.45% and 73.13% GT-known Loc, respectively. Even under the extreme setting of using only 1 image per class from ImageNet for training, SAT already exceeds the SOTA method by 2.1% GT-known Loc. Code and models are available at https://github.com/wpy1999/SAT.
Although current data augmentation methods are successful to alleviate the data insufficiency, conventional augmentation are primarily intra-domain while advanced generative adversarial networks (GANs) generate images remaining uncertain, particularly in small-scale datasets. In this paper, we propose a parameterized GAN (ParaGAN) that effectively controls the changes of synthetic samples among domains and highlights the attention regions for downstream classification. Specifically, ParaGAN incorporates projection distance parameters in cyclic projection and projects the source images to the decision boundary to obtain the class-difference maps. Our experiments show that ParaGAN can consistently outperform the existing augmentation methods with explainable classification on two small-scale medical datasets.
Image quality plays a big role in CNN-based image classification performance. Fine-tuning the network with distorted samples may be too costly for large networks. To solve this issue, we propose a transfer learning approach optimized to keep into account that in each layer of a CNN some filters are more susceptible to image distortion than others. Our method identifies the most susceptible filters and applies retraining only to the filters that show the highest activation maps distance between clean and distorted images. Filters are ranked using the Borda count election method and then only the most affected filters are fine-tuned. This significantly reduces the number of parameters to retrain. We evaluate this approach on the CIFAR-10 and CIFAR-100 datasets, testing it on two different models and two different types of distortion. Results show that the proposed transfer learning technique recovers most of the lost performance due to input data distortion, at a considerably faster pace with respect to existing methods, thanks to the reduced number of parameters to fine-tune. When few noisy samples are provided for training, our filter-level fine tuning performs particularly well, also outperforming state of the art layer-level transfer learning approaches.
Visual prompting (VP) is an emerging parameter-efficient fine-tuning approach to adapting pre-trained vision models to solve various downstream image-classification tasks. However, there has hitherto been little systematic study of the design space of VP and no clear benchmark for evaluating its performance. To bridge this gap, we propose AutoVP, an end-to-end expandable framework for automating VP design choices, along with 12 downstream image-classification tasks that can serve as a holistic VP-performance benchmark. Our design space covers 1) the joint optimization of the prompts; 2) the selection of pre-trained models, including image classifiers and text-image encoders; and 3) model output mapping strategies, including nonparametric and trainable label mapping. Our extensive experimental results show that AutoVP outperforms the best-known current VP methods by a substantial margin, having up to 6.7% improvement in accuracy; and attains a maximum performance increase of 27.5% compared to linear-probing (LP) baseline. AutoVP thus makes a two-fold contribution: serving both as an efficient tool for hyperparameter tuning on VP design choices, and as a comprehensive benchmark that can reasonably be expected to accelerate VP's development. The source code is available at https://github.com/IBM/AutoVP.
In this paper, we propose to employ the convolutional neural network (CNN) for the image question answering (QA). Our proposed CNN provides an end-to-end framework with convolutional architectures for learning not only the image and question representations, but also their inter-modal interactions to produce the answer. More specifically, our model consists of three CNNs: one image CNN to encode the image content, one sentence CNN to compose the words of the question, and one multimodal convolution layer to learn their joint representation for the classification in the space of candidate answer words. We demonstrate the efficacy of our proposed model on the DAQUAR and COCO-QA datasets, which are two benchmark datasets for the image QA, with the performances significantly outperforming the state-of-the-art.
In this work, we study over-parameterization as a necessary condition for having the ability for the models to extrapolate outside the convex hull of training set. We specifically, consider classification models, e.g., image classification and other applications of deep learning. Such models are classification functions that partition their domain and assign a class to each partition \cite{strang2019linear}. Partitions are defined by decision boundaries and so is the classification model/function. Convex hull of training set may occupy only a subset of the domain, but trained model may partition the entire domain and not just the convex hull of training set. This is important because many of the testing samples may be outside the convex hull of training set and the way in which a model partitions its domain outside the convex hull would be influential in its generalization. Using approximation theory, we prove that over-parameterization is a necessary condition for having control over the partitioning of the domain outside the convex hull of training set. We also propose a more clear definition for the notion of over-parametrization based on the learning task and the training set at hand. We present empirical evidence about geometry of datasets, both image and non-image, to provide insights about the extent of extrapolation performed by the models. We consider a 64-dimensional feature space learned by a ResNet model and investigate the geometric arrangements of convex hulls and decision boundaries in that space. We also formalize the notion of extrapolation and relate it to the scope of the model. Finally, we review the rich extrapolation literature in pure and applied mathematics, e.g., the Whitney's Extension Problem, and place our theory in that context.
Hierarchical classification is significant for complex tasks by providing multi-granular predictions and encouraging better mistakes. As the label structure decides its performance, many existing approaches attempt to construct an excellent label structure for promoting the classification results. In this paper, we consider that different label structures provide a variety of prior knowledge for category recognition, thus fusing them is helpful to achieve better hierarchical classification results. Furthermore, we propose a multi-task multi-structure fusion model to integrate different label structures. It contains two kinds of branches: one is the traditional classification branch to classify the common subclasses, the other is responsible for identifying the heterogeneous superclasses defined by different label structures. Besides the effect of multiple label structures, we also explore the architecture of the deep model for better hierachical classification and adjust the hierarchical evaluation metrics for multiple label structures. Experimental results on CIFAR100 and Car196 show that our method obtains significantly better results than using a flat classifier or a hierarchical classifier with any single label structure.