With the popularity of foundational models, parameter efficient fine tuning has become the defacto approach to leverage pretrained models to perform downstream tasks. Taking inspiration from recent advances in large language models, Visual Prompt Tuning, and similar techniques, learn an additional prompt to efficiently finetune a pretrained vision foundational model. However, we observe that such prompting is insufficient for fine-grained visual classification tasks such as medical image classification, where there is large inter-class variance, and small intra-class variance. Hence, in this paper we propose to leverage advanced segmentation capabilities of Segment Anything Model 2 (SAM2) as a visual prompting cue to help visual encoder in the CLIP (Contrastive Language-Image Pretraining) by guiding the attention in CLIP visual encoder to relevant regions in the image. This helps the model to focus on highly discriminative regions, without getting distracted from visually similar background features, an essential requirement in a fewshot, finegrained classification setting. We evaluate our method on diverse medical datasets including X-rays, CT scans, and MRI images, and report an accuracy of (71%, 81%, 86%, 58%) from the proposed approach on (COVID, lung-disease, brain-tumor, breast-cancer) datasets against (66%, 70%, 68%, 29%) from a pretrained CLIP model after fewshot training. The proposed approach also allows to obtain interpretable explanation for the classification performance through the localization obtained using segmentation.
Social media platforms often act as breeding grounds for various forms of trolling or malicious content targeting users or communities. One way of trolling users is by creating memes, which in most cases unites an image with a short piece of text embedded on top of it. The situation is more complex for multilingual(e.g., Tamil) memes due to the lack of benchmark datasets and models. We explore several models to detect Troll memes in Tamil based on the shared task, "Troll Meme Classification in DravidianLangTech2022" at ACL-2022. We observe while the text-based model MURIL performs better for Non-troll meme classification, the image-based model VGG16 performs better for Troll-meme classification. Further fusing these two modalities help us achieve stable outcomes in both classes. Our fusion model achieved a 0.561 weighted average F1 score and ranked second in this task.
Just like weights, bias terms are the learnable parameters of many popular machine learning models, including neural networks. Biases are thought to enhance the representational power of neural networks, enabling them to solve a variety of tasks in computer vision. However, we argue that biases can be disregarded for some image-related tasks such as image classification, by considering the intrinsic distribution of images in the input space and desired model properties from first principles. Our findings suggest that zero-bias neural networks can perform comparably to biased networks for practical image classification tasks. We demonstrate that zero-bias neural networks possess a valuable property called scalar (multiplication) invariance. This means that the prediction of the network remains unchanged when the contrast of the input image is altered. We extend scalar invariance to more general cases, enabling formal verification of certain convex regions of the input space. Additionally, we prove that zero-bias neural networks are fair in predicting the zero image. Unlike state-of-the-art models that may exhibit bias toward certain labels, zero-bias networks have uniform belief in all labels. We believe dropping bias terms can be considered as a geometric prior in designing neural network architecture for image classification, which shares the spirit of adapting convolutions as the transnational invariance prior. The robustness and fairness advantages of zero-bias neural networks may also indicate a promising path towards trustworthy and ethical AI.
Few-shot image classification has recently witnessed the rise of representation learning being utilised for models to adapt to new classes using only a few training examples. Therefore, the properties of the representations, such as their underlying probability distributions, assume vital importance. Representations sampled from Gaussian distributions have been used in recent works, [19] to train classifiers for few-shot classification. These methods rely on transforming the distributions of experimental data to approximate Gaussian distributions for their functioning. In this paper, I propose a novel Gaussian transform, that outperforms existing methods on transforming experimental data into Gaussian-like distributions. I then utilise this novel transformation for few-shot image classification and show significant gains in performance, while sampling lesser data.
The vigorous developments of Internet of Things make it possible to extend its computing and storage capabilities to computing tasks in the aerial system with collaboration of cloud and edge, especially for artificial intelligence (AI) tasks based on deep learning (DL). Collecting a large amount of image/video data, Unmanned aerial vehicles (UAVs) can only handover intelligent analysis tasks to the back-end mobile edge computing (MEC) server due to their limited storage and computing capabilities. How to efficiently transmit the most correlated information for the AI model is a challenging topic. Inspired by the task-oriented communication in recent years, we propose a new aerial image transmission paradigm for the scene classification task. A lightweight model is developed on the front-end UAV for semantic blocks transmission with perception of images and channel conditions. In order to achieve the tradeoff between transmission latency and classification accuracy, deep reinforcement learning (DRL) is used to explore the semantic blocks which have the best contribution to the back-end classifier under various channel conditions. Experimental results show that the proposed method can significantly improve classification accuracy compared to the fixed transmission strategy and traditional content perception methods.
Weakly supervised object localization (WSOL) relaxes the requirement of dense annotations for object localization by using image-level classification masks to supervise its learning process. However, current WSOL methods suffer from excessive activation of background locations and need post-processing to obtain the localization mask. This paper attributes these issues to the unawareness of background cues, and propose the background-aware classification activation map (B-CAM) to simultaneously learn localization scores of both object and background with only image-level labels. In our B-CAM, two image-level features, aggregated by pixel-level features of potential background and object locations, are used to purify the object feature from the object-related background and to represent the feature of the pure-background sample, respectively. Then based on these two features, both the object classifier and the background classifier are learned to determine the binary object localization mask. Our B-CAM can be trained in end-to-end manner based on a proposed stagger classification loss, which not only improves the objects localization but also suppresses the background activation. Experiments show that our B-CAM outperforms one-stage WSOL methods on the CUB-200, OpenImages and VOC2012 datasets.
Geometric shape features have been widely used as strong predictors for image classification. Nevertheless, most existing classifiers such as deep neural networks (DNNs) directly leverage the statistical correlations between these shape features and target variables. However, these correlations can often be spurious and unstable across different environments (e.g., in different age groups, certain types of brain changes have unstable relations with neurodegenerative disease); hence leading to biased or inaccurate predictions. In this paper, we introduce a novel framework that for the first time develops invariant shape representation learning (ISRL) to further strengthen the robustness of image classifiers. In contrast to existing approaches that mainly derive features in the image space, our model ISRL is designed to jointly capture invariant features in latent shape spaces parameterized by deformable transformations. To achieve this goal, we develop a new learning paradigm based on invariant risk minimization (IRM) to learn invariant representations of image and shape features across multiple training distributions/environments. By embedding the features that are invariant with regard to target variables in different environments, our model consistently offers more accurate predictions. We validate our method by performing classification tasks on both simulated 2D images, real 3D brain and cine cardiovascular magnetic resonance images (MRIs). Our code is publicly available at https://github.com/tonmoy-hossain/ISRL.
Height estimation has long been a pivotal topic within measurement and remote sensing disciplines, proving critical for endeavours such as 3D urban modelling, MR and autonomous driving. Traditional methods utilise stereo matching or multisensor fusion, both well-established techniques that typically necessitate multiple images from varying perspectives and adjunct sensors like SAR, leading to substantial deployment costs. Single image height estimation has emerged as an attractive alternative, boasting a larger data source variety and simpler deployment. However, current methods suffer from limitations such as fixed receptive fields, a lack of global information interaction, leading to noticeable instance-level height deviations. The inherent complexity of height prediction can result in a blurry estimation of object edge depth when using mainstream regression methods based on fixed height division. This paper presents a comprehensive solution for monocular height estimation in remote sensing, termed HeightFormer, combining multilevel interactions and image-adaptive classification-regression. It features the Multilevel Interaction Backbone (MIB) and Image-adaptive Classification-regression Height Generator (ICG). MIB supplements the fixed sample grid in CNN of the conventional backbone network with tokens of different interaction ranges. It is complemented by a pixel-, patch-, and feature map-level hierarchical interaction mechanism, designed to relay spatial geometry information across different scales and introducing a global receptive field to enhance the quality of instance-level height estimation. The ICG dynamically generates height partition for each image and reframes the traditional regression task, using a refinement from coarse to fine classification-regression that significantly mitigates the innate ill-posedness issue and drastically improves edge sharpness.
A Hyperspectral image contains much more number of channels as compared to a RGB image, hence containing more information about entities within the image. The convolutional neural network (CNN) and the Multi-Layer Perceptron (MLP) have been proven to be an effective method of image classification. However, they suffer from the issues of long training time and requirement of large amounts of the labeled data, to achieve the expected outcome. These issues become more complex while dealing with hyperspectral images. To decrease the training time and reduce the dependence on large labeled dataset, we propose using the method of transfer learning. The hyperspectral dataset is preprocessed to a lower dimension using PCA, then deep learning models are applied to it for the purpose of classification. The features learned by this model are then used by the transfer learning model to solve a new classification problem on an unseen dataset. A detailed comparison of CNN and multiple MLP architectural models is performed, to determine an optimum architecture that suits best the objective. The results show that the scaling of layers not always leads to increase in accuracy but often leads to overfitting, and also an increase in the training time.The training time is reduced to greater extent by applying the transfer learning approach rather than just approaching the problem by directly training a new model on large datasets, without much affecting the accuracy.
In this work, we present an end-to-end deep learning framework for X-ray image diagnosis. As the first step, our system determines whether a submitted image is an X-ray or not. After it classifies the type of the X-ray, it runs the dedicated abnormality classification network. In this work, we only focus on the chest X-rays for abnormality classification. However, the system can be extended to other X-ray types easily. Our deep learning classifiers are based on DenseNet-121 architecture. The test set accuracy obtained for 'X-ray or Not', 'X-ray Type Classification', and 'Chest Abnormality Classification' tasks are 0.987, 0.976, and 0.947, respectively, resulting into an end-to-end accuracy of 0.91. For achieving better results than the state-of-the-art in the 'Chest Abnormality Classification', we utilize the new RAdam optimizer. We also use Gradient-weighted Class Activation Mapping for visual explanation of the results. Our results show the feasibility of a generalized online projectional radiography diagnosis system.
We present a novel Tensor Composition Net (TCN) to predict visual relationships in images. Visual Relationship Prediction (VRP) provides a more challenging test of image understanding than conventional image tagging and is difficult to learn due to a large label-space and incomplete annotation. The key idea of our TCN is to exploit the low-rank property of the visual relationship tensor, so as to leverage correlations within and across objects and relations and make a structured prediction of all visual relationships in an image. To show the effectiveness of our model, we first empirically compare our model with Multi-Label Image Classification (MLIC) methods, eXtreme Multi-label Classification (XMC) methods, and VRD methods. We then show that thanks to our tensor (de)composition layer, our model can predict visual relationships which have not been seen in the training dataset. We finally show our TCN's image-level visual relationship prediction provides a simple and efficient mechanism for relation-based image-retrieval even compared with VRD methods.
The natural world is long-tailed: rare classes are observed orders of magnitudes less frequently than common ones, leading to highly-imbalanced data where rare classes can have only handfuls of examples. Learning from few examples is a known challenge for deep learning based classification algorithms, and is the focus of the field of low-shot learning. One potential approach to increase the training data for these rare classes is to augment the limited real data with synthetic samples. This has been shown to help, but the domain shift between real and synthetic hinders the approaches' efficacy when tested on real data. We explore the use of image-to-image translation methods to close the domain gap between synthetic and real imagery for animal species classification in data collected from camera traps: motion-activated static cameras used to monitor wildlife. We use low-level feature alignment between source and target domains to make synthetic data for a rare species generated using a graphics engine more "realistic". Compared against a system augmented with unaligned synthetic data, our experiments show a considerable decrease in classification error rates on a rare species.
The recent success of the generative model shows that leveraging the multi-modal embedding space can manipulate an image using text information. However, manipulating an image with other sources rather than text, such as sound, is not easy due to the dynamic characteristics of the sources. Especially, sound can convey vivid emotions and dynamic expressions of the real world. Here, we propose a framework that directly encodes sound into the multi-modal (image-text) embedding space and manipulates an image from the space. Our audio encoder is trained to produce a latent representation from an audio input, which is forced to be aligned with image and text representations in the multi-modal embedding space. We use a direct latent optimization method based on aligned embeddings for sound-guided image manipulation. We also show that our method can mix text and audio modalities, which enrich the variety of the image modification. We verify the effectiveness of our sound-guided image manipulation quantitatively and qualitatively. We also show that our method can mix different modalities, i.e., text and audio, which enrich the variety of the image modification. The experiments on zero-shot audio classification and semantic-level image classification show that our proposed model outperforms other text and sound-guided state-of-the-art methods.
Text classification is a fundamental task in NLP applications. Latest research in this field has largely been divided into two major sub-fields. Learning representations is one sub-field and learning deeper models, both sequential and convolutional, which again connects back to the representation is the other side. We posit the idea that the stronger the representation is, the simpler classifier models are needed to achieve higher performance. In this paper we propose a completely novel direction to text classification research, wherein we convert text to a representation very similar to images, such that any deep network able to handle images is equally able to handle text. We take a deeper look at the representation of documents as an image and subsequently utilize very simple convolution based models taken as is from computer vision domain. This image can be cropped, re-scaled, re-sampled and augmented just like any other image to work with most of the state-of-the-art large convolution based models which have been designed to handle large image datasets. We show impressive results with some of the latest benchmarks in the related fields. We perform transfer learning experiments, both from text to text domain and also from image to text domain. We believe this is a paradigm shift from the way document understanding and text classification has been traditionally done, and will drive numerous novel research ideas in the community.
Rotational motion blur caused by the circular motion of the camera or/and object is common in life. Identifying objects from images affected by rotational motion blur is challenging because this image degradation severely impacts image quality. Therefore, it is meaningful to develop image invariant features under rotational motion blur and then use them in practical tasks, such as object classification and template matching. This paper proposes a novel method to generate image moment invariants under general rotational motion blur and provides some instances. Further, we achieve their invariance to similarity transform. To the best of our knowledge, this is the first time that moment invariants for rotational motion blur have been proposed in the literature. We conduct extensive experiments on various image datasets disturbed by similarity transform and rotational motion blur to test these invariants' numerical stability and robustness to image noise. We also demonstrate their performance in image classification and handwritten digit recognition. Current state-of-the-art blur moment invariants and deep neural networks are chosen for comparison. Our results show that the moment invariants proposed in this paper significantly outperform other features in various tasks.
Convolutional neural networks (CNNs) have been recently used for a variety of histology image analysis. However, availability of a large dataset is a major prerequisite for training a CNN which limits its use by the computational pathology community. In previous studies, CNNs have demonstrated their potential in terms of feature generalizability and transferability accompanied with better performance. Considering these traits of CNN, we propose a simple yet effective method which leverages the strengths of CNN combined with the advantages of including contextual information, particularly designed for a small dataset. Our method consists of two main steps: first it uses the activation features of CNN trained for a patch-based classification and then it trains a separate classifier using features of overlapping patches to perform image-based classification using the contextual information. The proposed framework outperformed the state-of-the-art method for breast cancer classification.
Image representation and classification are two fundamental tasks towards multimedia content retrieval and understanding. The idea that shape and texture information (e.g. edge or orientation) are the key features for visual representation is ingrained and dominated in current multimedia and computer vision communities. A number of low-level features have been proposed by computing local gradients (e.g. SIFT, LBP and HOG), and have achieved great successes on numerous multimedia applications. In this paper, we present a simple yet efficient local descriptor for image classification, referred as Local Color Contrastive Descriptor (LCCD), by leveraging the neural mechanisms of color contrast. The idea originates from the observation in neural science that color and shape information are linked inextricably in visual cortical processing. The color contrast yields key information for visual color perception and provides strong linkage between color and shape. We propose a novel contrastive mechanism to compute the color contrast in both spatial location and multiple channels. The color contrast is computed by measuring \emph{f}-divergence between the color distributions of two regions. Our descriptor enriches local image representation with both color and contrast information. We verified experimentally that it can compensate strongly for the shape based descriptor (e.g. SIFT), while keeping computationally simple. Extensive experimental results on image classification show that our descriptor improves the performance of SIFT substantially by combinations, and achieves the state-of-the-art performance on three challenging benchmark datasets. It improves recent Deep Learning model (DeCAF) [1] largely from the accuracy of 40.94% to 49.68% in the large scale SUN397 database. Codes for the LCCD will be available.
The feature learning methods based on convolutional neural network (CNN) have successfully produced tremendous achievements in image classification tasks. However, the inherent noise and some other factors may weaken the effectiveness of the convolutional feature statistics. In this paper, we investigate Discrete Wavelet Transform (DWT) in the frequency domain and design a new Wavelet-Attention (WA) block to only implement attention in the high-frequency domain. Based on this, we propose a Wavelet-Attention convolutional neural network (WA-CNN) for image classification. Specifically, WA-CNN decomposes the feature maps into low-frequency and high-frequency components for storing the structures of the basic objects, as well as the detailed information and noise, respectively. Then, the WA block is leveraged to capture the detailed information in the high-frequency domain with different attention factors but reserves the basic object structures in the low-frequency domain. Experimental results on CIFAR-10 and CIFAR-100 datasets show that our proposed WA-CNN achieves significant improvements in classification accuracy compared to other related networks. Specifically, based on MobileNetV2 backbones, WA-CNN achieves 1.26% Top-1 accuracy improvement on the CIFAR-10 benchmark and 1.54% Top-1 accuracy improvement on the CIFAR-100 benchmark.
Ultrasound imaging is a commonly used technology for visualising patient anatomy in real-time during diagnostic and therapeutic procedures. High operator dependency and low reproducibility make ultrasound imaging and interpretation challenging with a steep learning curve. Automatic image classification using deep learning has the potential to overcome some of these challenges by supporting ultrasound training in novices, as well as aiding ultrasound image interpretation in patient with complex pathology for more experienced practitioners. However, the use of deep learning methods requires a large amount of data in order to provide accurate results. Labelling large ultrasound datasets is a challenging task because labels are retrospectively assigned to 2D images without the 3D spatial context available in vivo or that would be inferred while visually tracking structures between frames during the procedure. In this work, we propose a multi-modal convolutional neural network (CNN) architecture that labels endoscopic ultrasound (EUS) images from raw verbal comments provided by a clinician during the procedure. We use a CNN composed of two branches, one for voice data and another for image data, which are joined to predict image labels from the spoken names of anatomical landmarks. The network was trained using recorded verbal comments from expert operators. Our results show a prediction accuracy of 76% at image level on a dataset with 5 different labels. We conclude that the addition of spoken commentaries can increase the performance of ultrasound image classification, and eliminate the burden of manually labelling large EUS datasets necessary for deep learning applications.
Before the recent success of deep learning methods for automated medical image analysis, practitioners used handcrafted radiomic features to quantitatively describe local patches of medical images. However, extracting discriminative radiomic features relies on accurate pathology localization, which is difficult to acquire in real-world settings. Despite advances in disease classification and localization from chest X-rays, many approaches fail to incorporate clinically-informed domain knowledge. For these reasons, we propose a Radiomics-Guided Transformer (RGT) that fuses \textit{global} image information with \textit{local} knowledge-guided radiomics information to provide accurate cardiopulmonary pathology localization and classification \textit{without any bounding box annotations}. RGT consists of an image Transformer branch, a radiomics Transformer branch, and fusion layers that aggregate image and radiomic information. Using the learned self-attention of its image branch, RGT extracts a bounding box for which to compute radiomic features, which are further processed by the radiomics branch; learned image and radiomic features are then fused and mutually interact via cross-attention layers. Thus, RGT utilizes a novel end-to-end feedback loop that can bootstrap accurate pathology localization only using image-level disease labels. Experiments on the NIH ChestXRay dataset demonstrate that RGT outperforms prior works in weakly supervised disease localization (by an average margin of 3.6\% over various intersection-over-union thresholds) and classification (by 1.1\% in average area under the receiver operating characteristic curve). We publicly release our codes and pre-trained models at \url{https://github.com/VITA-Group/chext}.