Video Relationship Detection Using Mixture of Experts

Machine comprehension of visual information from images and videos by neural networks faces two primary challenges. Firstly, there exists a computational and inference gap in connecting vision and language, making it difficult to accurately determine which object a given agent acts on and represent it through language. Secondly, classifiers trained by a single, monolithic neural network often lack stability and generalization. To overcome these challenges, we introduce MoE-VRD, a novel approach to visual relationship detection utilizing a mixture of experts. MoE-VRD identifies language triplets in the form of < subject, predicate, object> tuples to extract relationships from visual processing. Leveraging recent advancements in visual relationship detection, MoE-VRD addresses the requirement for action recognition in establishing relationships between subjects (acting) and objects (being acted upon). In contrast to single monolithic networks, MoE-VRD employs multiple small models as experts, whose outputs are aggregated. Each expert in MoE-VRD specializes in visual relationship learning and object tagging. By utilizing a sparsely-gated mixture of experts, MoE-VRD enables conditional computation and significantly enhances neural network capacity without increasing computational complexity. Our experimental results demonstrate that the conditional computation capabilities and scalability of the mixture-of-experts approach lead to superior performance in visual relationship detection compared to state-of-the-art methods.

EPNet: An Efficient Pyramid Network for Enhanced Single-Image Super-Resolution with Reduced Computational Requirements

Single-image super-resolution (SISR) has seen significant advancements through the integration of deep learning. However, the substantial computational and memory requirements of existing methods often limit their practical application. This paper introduces a new Efficient Pyramid Network (EPNet) that harmoniously merges an Edge Split Pyramid Module (ESPM) with a Panoramic Feature Extraction Module (PFEM) to overcome the limitations of existing methods, particularly in terms of computational efficiency. The ESPM applies a pyramid-based channel separation strategy, boosting feature extraction while maintaining computational efficiency. The PFEM, a novel fusion of CNN and Transformer structures, enables the concurrent extraction of local and global features, thereby providing a panoramic view of the image landscape. Our architecture integrates the PFEM in a manner that facilitates the streamlined exchange of feature information and allows for the further refinement of image texture details. Experimental results indicate that our model outperforms existing state-of-the-art methods in image resolution quality, while considerably decreasing computational and memory costs. This research contributes to the ongoing evolution of efficient and practical SISR methodologies, bearing broader implications for the field of computer vision.

Challenges for Predictive Modeling with Neural Network Techniques using Error-Prone Dietary Intake Data

Dietary intake data are routinely drawn upon to explore diet-health relationships. However, these data are often subject to measurement error, distorting the true relationships. Beyond measurement error, there are likely complex synergistic and sometimes antagonistic interactions between different dietary components, complicating the relationships between diet and health outcomes. Flexible models are required to capture the nuance that these complex interactions introduce. This complexity makes research on diet-health relationships an appealing candidate for the application of machine learning techniques, and in particular, neural networks. Neural networks are computational models that are able to capture highly complex, nonlinear relationships so long as sufficient data are available. While these models have been applied in many domains, the impacts of measurement error on the performance of predictive modeling has not been systematically investigated. However, dietary intake data are typically collected using self-report methods and are prone to large amounts of measurement error. In this work, we demonstrate the ways in which measurement error erodes the performance of neural networks, and illustrate the care that is required for leveraging these models in the presence of error. We demonstrate the role that sample size and replicate measurements play on model performance, indicate a motivation for the investigation of transformations to additivity, and illustrate the caution required to prevent model overfitting. While the past performance of neural networks across various domains make them an attractive candidate for examining diet-health relationships, our work demonstrates that substantial care and further methodological development are both required to observe increased predictive performance when applying these techniques, compared to more traditional statistical procedures.

Memory-Efficient Continual Learning Object Segmentation for Long Video

Recent state-of-the-art semi-supervised Video Object Segmentation (VOS) methods have shown significant improvements in target object segmentation accuracy when information from preceding frames is used in undertaking segmentation on the current frame. In particular, such memory-based approaches can help a model to more effectively handle appearance changes (representation drift) or occlusions. Ideally, for maximum performance, online VOS methods would need all or most of the preceding frames (or their extracted information) to be stored in memory and be used for online learning in consecutive frames. Such a solution is not feasible for long videos, as the required memory size would grow without bound. On the other hand, these methods can fail when memory is limited and a target object experiences repeated representation drifts throughout a video. We propose two novel techniques to reduce the memory requirement of online VOS methods while improving modeling accuracy and generalization on long videos. Motivated by the success of continual learning techniques in preserving previously-learned knowledge, here we propose Gated-Regularizer Continual Learning (GRCL), which improves the performance of any online VOS subject to limited memory, and a Reconstruction-based Memory Selection Continual Learning (RMSCL) which empowers online VOS methods to efficiently benefit from stored information in memory. Experimental results show that the proposed methods improve the performance of online VOS models up to 10 %, and boosts their robustness on long-video datasets while maintaining comparable performance on short-video datasets DAVIS16 and DAVIS17.

A Step Towards Worldwide Biodiversity Assessment: The BIOSCAN-1M Insect Dataset

In an effort to catalog insect biodiversity, we propose a new large dataset of hand-labelled insect images, the BIOSCAN-Insect Dataset. Each record is taxonomically classified by an expert, and also has associated genetic information including raw nucleotide barcode sequences and assigned barcode index numbers, which are genetically-based proxies for species classification. This paper presents a curated million-image dataset, primarily to train computer-vision models capable of providing image-based taxonomic assessment, however, the dataset also presents compelling characteristics, the study of which would be of interest to the broader machine learning community. Driven by the biological nature inherent to the dataset, a characteristic long-tailed class-imbalance distribution is exhibited. Furthermore, taxonomic labelling is a hierarchical classification scheme, presenting a highly fine-grained classification problem at lower levels. Beyond spurring interest in biodiversity research within the machine learning community, progress on creating an image-based taxonomic classifier will also further the ultimate goal of all BIOSCAN research: to lay the foundation for a comprehensive survey of global biodiversity. This paper introduces the dataset and explores the classification task through the implementation and analysis of a baseline classifier.

Multi-Channel Feature Extraction for Virtual Histological Staining of Photon Absorption Remote Sensing Images

Accurate and fast histological staining is crucial in histopathology, impacting diagnostic precision and reliability. Traditional staining methods are time-consuming and subjective, causing delays in diagnosis. Digital pathology plays a vital role in advancing and optimizing histology processes to improve efficiency and reduce turnaround times. This study introduces a novel deep learning-based framework for virtual histological staining using photon absorption remote sensing (PARS) images. By extracting features from PARS time-resolved signals using a variant of the K-means method, valuable multi-modal information is captured. The proposed multi-channel cycleGAN (MC-GAN) model expands on the traditional cycleGAN framework, allowing the inclusion of additional features. Experimental results reveal that specific combinations of features outperform the conventional channels by improving the labeling of tissue structures prior to model training. Applied to human skin and mouse brain tissue, the results underscore the significance of choosing the optimal combination of features, as it reveals a substantial visual and quantitative concurrence between the virtually stained and the gold standard chemically stained hematoxylin and eosin (H&E) images, surpassing the performance of other feature combinations. Accurate virtual staining is valuable for reliable diagnostic information, aiding pathologists in disease classification, grading, and treatment planning. This study aims to advance label-free histological imaging and opens doors for intraoperative microscopy applications.

Virtual Histology with Photon Absorption Remote Sensing using a Cycle-Consistent Generative Adversarial Network with Weakly Registered Pairs

Modern histopathology relies on the microscopic examination of thin tissue sections stained with histochemical techniques, typically using brightfield or fluorescence microscopy. However, the staining of samples can permanently alter their chemistry and structure, meaning an individual tissue section must be prepared for each desired staining contrast. This not only consumes valuable tissue samples but also introduces delays in essential diagnostic timelines. In this work, virtual histochemical staining is developed using label-free photon absorption remote sensing (PARS) microscopy. We present a method that generates virtually stained histology images that are indistinguishable from the gold standard hematoxylin and eosin (H&E) staining. First, PARS label-free ultraviolet absorption images are captured directly within unstained tissue specimens. The radiative and non-radiative absorption images are then preprocessed, and virtually stained through the presented pathway. The preprocessing pipeline features a self-supervised Noise2Void denoising convolutional neural network (CNN) as well as a novel algorithm for pixel-level mechanical scanning error correction. These developments significantly enhance the recovery of sub-micron tissue structures, such as nucleoli location and chromatin distribution. Finally, we used a cycle-consistent generative adversarial network CycleGAN architecture to virtually stain the preprocessed PARS data. Virtual staining is applied to thin unstained sections of malignant human skin and breast tissue samples. Clinically relevant details are revealed, with comparable contrast and quality to gold standard H&E-stained images. This work represents a crucial step to deploying label-free microscopy as an alternative to standard histopathology techniques.

Is Generative Modeling-based Stylization Necessary for Domain Adaptation in Regression Tasks?

Unsupervised domain adaptation (UDA) aims to bridge the gap between source and target domains in the absence of target domain labels using two main techniques: input-level alignment (such as generative modeling and stylization) and feature-level alignment (which matches the distribution of the feature maps, e.g. gradient reversal layers). Motivated from the success of generative modeling for image classification, stylization-based methods were recently proposed for regression tasks, such as pose estimation. However, use of input-level alignment via generative modeling and stylization incur additional overhead and computational complexity which limit their use in real-world DA tasks. To investigate the role of input-level alignment for DA, we ask the following question: Is generative modeling-based stylization necessary for visual domain adaptation in regression? Surprisingly, we find that input-alignment has little effect on regression tasks as compared to classification. Based on these insights, we develop a non-parametric feature-level domain alignment method -- Implicit Stylization (ImSty) -- which results in consistent improvements over SOTA regression task, without the need for computationally intensive stylization and generative modeling. Our work conducts a critical evaluation of the role of generative modeling and stylization, at a time when these are also gaining popularity for domain generalization.

CLVOS23: A Long Video Object Segmentation Dataset for Continual Learning

Continual learning in real-world scenarios is a major challenge. A general continual learning model should have a constant memory size and no predefined task boundaries, as is the case in semi-supervised Video Object Segmentation (VOS), where continual learning challenges particularly present themselves in working on long video sequences. In this article, we first formulate the problem of semi-supervised VOS, specifically online VOS, as a continual learning problem, and then secondly provide a public VOS dataset, CLVOS23, focusing on continual learning. Finally, we propose and implement a regularization-based continual learning approach on LWL, an existing online VOS baseline, to demonstrate the efficacy of continual learning when applied to online VOS and to establish a CLVOS23 baseline. We apply the proposed baseline to the Long Videos dataset as well as to two short video VOS datasets, DAVIS16 and DAVIS17. To the best of our knowledge, this is the first time that VOS has been defined and addressed as a continual learning problem.