Despite the deep neural networks (DNN) has achieved excellent performance in image classification researches, the training of DNNs needs a large of clean data with accurate annotations. The collect of a dataset is easy, but it is difficult to annotate the collecting data. On the websites, there exist a lot of image data which contains inaccurate annotations, but training on these datasets may make networks easier to over-fit the noisy labels and cause performance degradation. In this work, we propose an improved joint optimization framework, which mixed the mix-up entropy and Kullback-Leibler (KL) entropy as the loss function. The new loss function can give the better fine-tuning after the framework updates both the label annotations. We conduct experiments on CIFAR-10 dataset and Clothing1M dataset. The result shows the advantageous performance of our approach compared with other state-of-the-art methods.
Convolutional neural networks (CNNs) are commonly used for image classification tasks, raising the challenge of their application on data flows. During their training, adaptation is often performed by tuning the learning rate. Usual learning rate strategies are time-based i.e. monotonously decreasing. In this paper, we advocate switching to a performance-based adaptation, in order to improve the learning efficiency. We present E (Exponential)/PD (Proportional Derivative)-Control, a conditional learning rate strategy that combines a feedback PD controller based on the CNN loss function, with an exponential control signal to smartly boost the learning and adapt the PD parameters. Stability proof is provided as well as an experimental evaluation using two state of the art image datasets (CIFAR-10 and Fashion-MNIST). Results show better performances than the related works (faster network accuracy growth reaching higher levels) and robustness of the E/PD-Control regarding its parametrization.
We consider the problem of unsupervised domain adaptation for image classification. To learn target-domain-aware features from the unlabeled data, we create a self-supervised pretext task by augmenting the unlabeled data with a certain type of transformation (specifically, image rotation) and ask the learner to predict the properties of the transformation. However, the obtained feature representation may contain a large amount of irrelevant information with respect to the main task. To provide further guidance, we force the feature representation of the augmented data to be consistent with that of the original data. Intuitively, the consistency introduces additional constraints to representation learning, therefore, the learned representation is more likely to focus on the right information about the main task. Our experimental results validate the proposed method and demonstrate state-of-the-art performance on classical domain adaptation benchmarks. Code is available at https://github.com/Jiaolong/ss-da-consistency.
We report results on unsupervised organization of cervical cells using microscopy of Pap-smear samples in brightfield (3-channel colour) as well as high resolution quantitative phase imaging modalities. A number of morphological parameters are measured for each of the 1450 cell nuclei (from $10$ woman subjects) imaged in this study. The Principal Component Analysis (PCA) methodology applied to this data shows that the cell image clustering performance improves significantly when brightfield as well as phase information is utilized for PCA as compared to when brightfield-only information is used. The results point to the feasibility of an image-based tool that will be able to mark suspicious cells for further examination by the pathologist. More importantly our results suggest that the information in quantitative phase images of cells that is typically not used in clinical practice is valuable for automated cell classification applications in general.
Quality assessment of agricultural produce is a crucial step in minimizing food stock wastage. However, this is currently done manually and often requires expert supervision, especially in smaller seeds like corn. We propose a novel computer vision-based system for automating this process. We build a novel seed image acquisition setup, which captures both the top and bottom views. Dataset collection for this problem has challenges of data annotation costs/time and class imbalance. We address these challenges by i.) using a Conditional Generative Adversarial Network (CGAN) to generate real-looking images for the classes with lesser images and ii.) annotate a large dataset with minimal expert human intervention by using a Batch Active Learning (BAL) based annotation tool. We benchmark different image classification models on the dataset obtained. We are able to get accuracies of up to 91.6% for testing the physical purity of seed samples.
Pre-training is essential to deep learning model performance, especially in medical image analysis tasks where limited training data are available. However, existing pre-training methods are inflexible as the pre-trained weights of one model cannot be reused by other network architectures. In this paper, we propose an architecture-irrelevant hyper-initializer, which can initialize any given network architecture well after being pre-trained for only once. The proposed initializer is a hypernetwork which takes a downstream architecture as input graphs and outputs the initialization parameters of the respective architecture. We show the effectiveness and efficiency of the hyper-initializer through extensive experimental results on multiple medical imaging modalities, especially in data-limited fields. Moreover, we prove that the proposed algorithm can be reused as a favorable plug-and-play initializer for any downstream architecture and task (both classification and segmentation) of the same modality.
Image-based machine learning models can be used to make the sorting and grading of agricultural products more efficient. In many regions, implementing such systems can be difficult due to the lack of centralization and automation of postharvest supply chains. Stakeholders are often too small to specialize in machine learning, and large training data sets are unavailable. We propose a machine learning procedure for images based on pre-trained Vision Transformers. It is easier to implement than the current standard approach of training Convolutional Neural Networks (CNNs) as we do not (re-)train deep neural networks. We evaluate our approach based on two data sets for apple defect detection and banana ripeness estimation. Our model achieves a competitive classification accuracy equal to or less than one percent below the best-performing CNN. At the same time, it requires three times fewer training samples to achieve a 90% accuracy.
Task transfer learning is a popular technique in image processing applications that uses pre-trained models to reduce the supervision cost of related tasks. An important question is to determine task transferability, i.e. given a common input domain, estimating to what extent representations learned from a source task can help in learning a target task. Typically, transferability is either measured experimentally or inferred through task relatedness, which is often defined without a clear operational meaning. In this paper, we present a novel metric, H-score, an easily-computable evaluation function that estimates the performance of transferred representations from one task to another in classification problems using statistical and information theoretic principles. Experiments on real image data show that our metric is not only consistent with the empirical transferability measurement, but also useful to practitioners in applications such as source model selection and task transfer curriculum learning.
Online Social Networks have revolutionized how we consume and share information, but they have also led to a proliferation of content not always reliable and accurate. One particular type of social accounts is known to promote unreputable content, hyperpartisan, and propagandistic information. They are automated accounts, commonly called bots. Focusing on Twitter accounts, we propose a novel approach to bot detection: we first propose a new algorithm that transforms the sequence of actions that an account performs into an image; then, we leverage the strength of Convolutional Neural Networks to proceed with image classification. We compare our performances with state-of-the-art results for bot detection on genuine accounts / bot accounts datasets well known in the literature. The results confirm the effectiveness of the proposal, because the detection capability is on par with the state of the art, if not better in some cases.
3D open-vocabulary scene graph methods are a promising map representation for embodied agents, however many current approaches are computationally expensive. In this paper, we reexamine the critical design choices established in previous works to optimize both efficiency and performance. We propose a general scene graph framework and conduct three studies that focus on image pre-processing, feature fusion, and feature selection. Our findings reveal that commonly used image pre-processing techniques provide minimal performance improvement while tripling computation (on a per object view basis). We also show that averaging feature labels across different views significantly degrades performance. We study alternative feature selection strategies that enhance performance without adding unnecessary computational costs. Based on our findings, we introduce a computationally balanced approach for 3D point cloud segmentation with per-object features. The approach matches state-of-the-art classification accuracy while achieving a threefold reduction in computation.
단독으로는 분류하기 어려운 이미지도 같은 비라벨 배치 안에 도움이 되는 이웃을 가지고 있을 수 있다. 제로샷 비전-언어 모델(VLM)은 이미지를 독립적으로 점수화하므로 이러한 증거를 활용하지 못한다. 본 연구는 라벨이 있는 예시가 전혀 없는 소규모 비라벨 질의 배치를 다룬다. 시각적 유사도는 자세나 배경을 반영할 수 있어 서로 다른 클래스의 이미지를 연결할 수 있다. 클래스 어휘 전반에 대한 각 이미지의 점수는 이웃을 선택하는 또 다른 방법을 제공한다. 본 논문에서는 이러한 클래스 응답 프로파일로부터 희소 그래프를 구축하는 행동 친화도 트랜스덕션(Behavioral Affinity Transduction, BAT)을 제안한다. 이 그래프는 예측 과정에서 이웃 증거를 제공한다. EM-Dirichlet의 경우 후보 클래스를 걸러내고 할당을 초기화하는 데에도 사용된다. BAT는 고정된 VLM의 직접 방식 및 반복 방식 리드아웃에 모두 적용할 수 있다. 다양한 데이터셋에서 BAT는 시험한 모든 리드아웃의 평균 정확도를 향상시킨다. 주된 RN50 비교에서 소프트 및 하드 EM-Dirichlet은 각각 $4.22$ 및 $2.72$ 퍼센트포인트 향상되었다.
Few-shot learning aims to recognize novel classes from a few examples. Although significant progress has been made in the image domain, few-shot video classification is relatively unexplored. We argue that previous methods underestimate the importance of video feature learning and propose to learn spatiotemporal features using a 3D CNN. Proposing a two-stage approach that learns video features on base classes followed by fine-tuning the classifiers on novel classes, we show that this simple baseline approach outperforms prior few-shot video classification methods by over 20 points on existing benchmarks. To circumvent the need of labeled examples, we present two novel approaches that yield further improvement. First, we leverage tag-labeled videos from a large dataset using tag retrieval followed by selecting the best clips with visual similarities. Second, we learn generative adversarial networks that generate video features of novel classes from their semantic embeddings. Moreover, we find existing benchmarks are limited because they only focus on 5 novel classes in each testing episode and introduce more realistic benchmarks by involving more novel classes, i.e. few-shot learning, as well as a mixture of novel and base classes, i.e. generalized few-shot learning. The experimental results show that our retrieval and feature generation approach significantly outperform the baseline approach on the new benchmarks.
Machine Learning (ML) has achieved enormous success in solving a variety of problems in computer vision, speech recognition, object detection, to name a few. The principal reason for this success is the availability of huge datasets for training deep neural networks (DNNs). However, datasets can not be publicly released if they contain sensitive information such as medical or financial records. In such cases, data privacy becomes a major concern. Encryption methods offer a possible solution to this issue, however their deployment on ML applications is non-trivial, as they seriously impact the classification accuracy and result in substantial computational overhead.Alternatively, obfuscation techniques can be used, but maintaining a good balance between visual privacy and accuracy is challenging. In this work, we propose a method to generate secure synthetic datasets from the original private datasets. In our method, given a network with Batch Normalization (BN) layers pre-trained on the original dataset, we first record the layer-wise BN statistics. Next, using the BN statistics and the pre-trained model, we generate the synthetic dataset by optimizing random noises such that the synthetic data match the layer-wise statistical distribution of the original model. We evaluate our method on image classification dataset (CIFAR10) and show that our synthetic data can be used for training networks from scratch, producing reasonable classification performance.
In few-shot unsupervised domain adaptation (FS-UDA), most existing methods followed the few-shot learning (FSL) methods to leverage the low-level local features (learned from conventional convolutional models, e.g., ResNet) for classification. However, the goal of FS-UDA and FSL are relevant yet distinct, since FS-UDA aims to classify the samples in target domain rather than source domain. We found that the local features are insufficient to FS-UDA, which could introduce noise or bias against classification, and not be used to effectively align the domains. To address the above issues, we aim to refine the local features to be more discriminative and relevant to classification. Thus, we propose a novel task-specific semantic feature learning method (TSECS) for FS-UDA. TSECS learns high-level semantic features for image-to-class similarity measurement. Based on the high-level features, we design a cross-domain self-training strategy to leverage the few labeled samples in source domain to build the classifier in target domain. In addition, we minimize the KL divergence of the high-level feature distributions between source and target domains to shorten the distance of the samples between the two domains. Extensive experiments on DomainNet show that the proposed method significantly outperforms SOTA methods in FS-UDA by a large margin (i.e., 10%).
This paper develops the MUFIN technique for extreme classification (XC) tasks with millions of labels where datapoints and labels are endowed with visual and textual descriptors. Applications of MUFIN to product-to-product recommendation and bid query prediction over several millions of products are presented. Contemporary multi-modal methods frequently rely on purely embedding-based methods. On the other hand, XC methods utilize classifier architectures to offer superior accuracies than embedding only methods but mostly focus on text-based categorization tasks. MUFIN bridges this gap by reformulating multi-modal categorization as an XC problem with several millions of labels. This presents the twin challenges of developing multi-modal architectures that can offer embeddings sufficiently expressive to allow accurate categorization over millions of labels; and training and inference routines that scale logarithmically in the number of labels. MUFIN develops an architecture based on cross-modal attention and trains it in a modular fashion using pre-training and positive and negative mining. A novel product-to-product recommendation dataset MM-AmazonTitles-300K containing over 300K products was curated from publicly available amazon.com listings with each product endowed with a title and multiple images. On the all datasets MUFIN offered at least 3% higher accuracy than leading text-based, image-based and multi-modal techniques. Code for MUFIN is available at https://github.com/Extreme-classification/MUFIN
This paper deals with the convergence analysis of the SUCPA (Semi Unsupervised Calibration through Prior Adaptation) algorithm, defined from a first-order non-linear difference equations, first developed to correct the scores output by a supervised machine learning classifier. The convergence analysis is addressed as a dynamical system problem, by studying the local and global stability of the nonlinear map derived from the algorithm. This map, which is defined by a composition of exponential and rational functions, turns out to be non-hyperbolic with a non-bounded set of non-isolated fixed points. Hence, a non-standard method for solving the convergence analysis is used consisting of an ad-hoc geometrical approach. For a binary classification problem (two-dimensional map), we rigorously prove that the map is globally asymptotically stable. Numerical experiments on real-world application are performed to support the theoretical results by means of two different classification problems: Sentiment Polarity performed with a Large Language Model and Cat-Dog Image classification. For a greater number of classes, the numerical evidence shows the same behavior of the algorithm, and this is illustrated with a Natural Language Inference example. The experiment codes are publicly accessible online at the following repository: https://github.com/LautaroEst/sucpa-convergence
병리학 파운데이션 모델(FM)은 일반적으로 라벨이 지정되지 않은 방대한 양의 데이터로 학습(training)된 모델로, 다운스트림 분류(classification) 작업에 효과적으로 사용할 수 있는 정규화된 잠재 공간(latent space)을 생성하는 것으로 나타났습니다. 이는 유사분열상(mitotic figure)과 기타 세포의 분류에서도 마찬가지입니다. 그러나 현재 FM의 잠재 공간이 밀집 객체 검출(dense object detection) 패러다임의 백본으로도 기능할 수 있을 만큼 판별력 있고 공간적으로 적절히 분해능을 갖춘 특징(feature)을 제공하는지는 지금까지 불분명합니다. 본 연구에서는 널리 사용되는 현재의 병리학 FM(UNI, UNI2-h, Virchow, Virchow2, H-optimus-0, H-optimus-1)에 대해 이 문제를 조사하고, ResNet50 아키텍처 기반의 완전한 엔드투엔드(end-to-end) 학습 기준선(baseline)과 성능을 비교합니다. 다중 도메인 MIDOG++ 데이터셋(dataset)과 도메인 외(out-of-domain) 사례인 TUPAC16 데이터셋에서 FM 백본을 단일 단계, 2단계, 자체 어텐션(self-attention) 기반 검출기의 대표 모델(각각 RetinaNet, Faster R-CNN, Deformable DETR)과 결합합니다. 실험 결과 H-optimus-0 및 Virchow 모델이 경쟁력 있는 성능을 보였으며, 이는 모두 이미지 수준의 자기지도학습(self-supervision)으로 학습된 현재 FM의 잠재 공간이 직접적인 유사분열상 검출에 적합하며 도메인 외 테스트 사례에서 약간 더 강건할 수 있음을 나타냅니다. 모든 코드는 https://github.com/DeepMicroscopy/FM4MFdet 에서 공개적으로 이용할 수 있습니다.
Current object detectors are limited in vocabulary size due to the small scale of detection datasets. Image classifiers, on the other hand, reason about much larger vocabularies, as their datasets are larger and easier to collect. We propose Detic, which simply trains the classifiers of a detector on image classification data and thus expands the vocabulary of detectors to tens of thousands of concepts. Unlike prior work, Detic does not need complex assignment schemes to assign image labels to boxes based on model predictions, making it much easier to implement and compatible with a range of detection architectures and backbones. Our results show that Detic yields excellent detectors even for classes without box annotations. It outperforms prior work on both open-vocabulary and long-tail detection benchmarks. Detic provides a gain of 2.4 mAP for all classes and 8.3 mAP for novel classes on the open-vocabulary LVIS benchmark. On the standard LVIS benchmark, Detic obtains 41.7 mAP when evaluated on all classes, or only rare classes, hence closing the gap in performance for object categories with few samples. For the first time, we train a detector with all the twenty-one-thousand classes of the ImageNet dataset and show that it generalizes to new datasets without finetuning. Code is available at \url{https://github.com/facebookresearch/Detic}.
Convolutional Neural Networks (CNNs) have been successful in solving tasks in computer vision including medical image segmentation due to their ability to automatically extract features from unstructured data. However, CNNs are sensitive to rotation and affine transformation and their success relies on huge-scale labeled datasets capturing various input variations. This network paradigm has posed challenges at scale because acquiring annotated data for medical segmentation is expensive, and strict privacy regulations. Furthermore, visual representation learning with CNNs has its own flaws, e.g., it is arguable that the pooling layer in traditional CNNs tends to discard positional information and CNNs tend to fail on input images that differ in orientations and sizes. Capsule network (CapsNet) is a recent new architecture that has achieved better robustness in representation learning by replacing pooling layers with dynamic routing and convolutional strides, which has shown potential results on popular tasks such as classification, recognition, segmentation, and natural language processing. Different from CNNs, which result in scalar outputs, CapsNet returns vector outputs, which aim to preserve the part-whole relationships. In this work, we first introduce the limitations of CNNs and fundamentals of CapsNet. We then provide recent developments of CapsNet for the task of medical image segmentation. We finally discuss various effective network architectures to implement a CapsNet for both 2D images and 3D volumetric medical image segmentation.
The COVID-19 can cause severe pneumonia and is estimated to have a high impact on the healthcare system. The standard image diagnosis tests for pneumonia are chest X-ray (CXR) and computed tomography (CT) scan. CXR are useful in because it is cheaper, faster and more widespread than CT. This study aims to identify pneumonia caused by COVID-19 from other types and also healthy lungs using only CXR images. In order to achieve the objectives, we have proposed a classification schema considering the multi-class and hierarchical perspectives, since pneumonia can be structured as a hierarchy. Given the natural data imbalance in this domain, we also proposed the use of resampling algorithms in order to re-balance the classes distribution. Our classification schema extract features using some well-known texture descriptors and also using a pre-trained CNN model. We also explored early and late fusion techniques in order to leverage the strength of multiple texture descriptors and base classifiers at once. To evaluate the approach, we composed a database, named RYDLS-20, containing CXR images of pneumonia caused by different pathogens as well as CXR images of healthy lungs. The classes distribution follows a real-world scenario in which some pathogens are more common than others. The proposed approach achieved a macro-avg F1-Score of 0.65 using a multi-class approach and a F1-Score of 0.89 for the COVID-19 identification in the hierarchical classification scenario. As far as we know, we achieved the best nominal rate obtained for COVID-19 identification in an unbalanced environment with more than three classes. We must also highlight the novel proposed hierarchical classification approach for this task, which considers the types of pneumonia caused by the different pathogens and lead us to the best COVID-19 recognition rate obtained here.