Retrieving images in response to textual queries requires some knowledge of the semantics picture. Here, we show how can do both automatic image annotation and retrieval (using one word queries) from videos using a multiple Bernoulli relevance model. The model assumes that training set or along with keyword annotations is provided. Multiple keywords are provided for an specific correspondence between not Each partitioned into rectangular regions real-valued feature vector computed over these...

Federated learning (FL) is a privacy-preserving paradigm for training collective machine models with locally stored data from multiple participants. Vertical federated (VFL) deals the case where participants sharing same sample ID space but having different feature spaces, while label information owned by one participant. Current studies of VFL only support two participants, and mostly focus on binaryclass logistic regression problems. In this paper, we propose Multi-participant Multi-class...

High spectral resolution brings hyperspectral images with large amounts of information, which makes these more useful in many applications than obtained from traditional multispectral scanners low resolution. However, the high data dimensionality increases burden on computation, storage, and transmission; fortunately, redundancy domain allows for significant reduction. Band selection provides a simple reduction scheme by discarding bands that are highly redundant, thereby preserving...

Federated learning (FL) is a privacy-preserving collective machine paradigm. Vertical federated (VFL) deals with the case where participants share same sample ID space but have different feature spaces, while label information owned by one participant. Early studies of VFL supported two and focused on binary-class logistic regression problems, recent put more attention specific aspects such as communication efficiency data security. In this paper, we propose multi-participant multi-class...

We apply a continuous relevance model (CRM) to the problem of directly retrieving visual content videos using text queries. The computes joint probability for image features and words training set annotated images. may then be used annotate unseen test probabilistic annotations are retrieval also propose modified - normalized CRM which substantially improves performance on subset TREC video dataset.

A brain-computer interface (BCI) based on code modulated visual evoked potentials (c-VEP) is among the fastest BCIs that have ever been reported, but it has not yet given a thorough study. In this study, pseudorandom binary M sequence and its time lag sequences are utilized for modulation of different stimuli template matching adopted as method target recognition. Five experiments were devised to investigate effect stimulus specificity recognition we made an effort find optimal parameters...

Nuclear instance segmentation and classification provide critical quantitative foundations for digital pathology diagnosis. With the advent of foundational Segment Anything Model (SAM), accuracy efficiency nuclear have improved significantly. However, SAM imposes a strong reliance on precise prompts, its class-agnostic design renders results entirely dependent provided prompts. Therefore, we focus generating prompts with more accurate localization propose \textbf{APSeg},...

In this paper we describe a novel approach for jointly modeling the text and visual components of multimedia documents purpose information retrieval(IR). We propose framework where individual are developed to model different relationships between queries then combined into joint retrieval framework. state-of-the-art systems, late combination two independent one analyzing just part such documents, other without leveraging any knowledge acquired in processing, is norm. Such systems rarely...

For tabletop object rearrangement problems with overhand grasps, storage space which may be inside or outside the workspace, running buffers, can temporarily hold objects greatly facilitates resolution of a given task.This brings forth natural question how many buffers are required so that certain classes feasible.In this work, we examine problem for both labeled (where each has specific goal pose) and unlabeled poses interchangeable) settings.On structural side, observe finding minimum...

Computational understanding of humor is an important topic under creative language and modeling. It can play a key role in complex human-AI interactions. The challenge here that human perception humorous content highly subjective. same joke may receive different funniness ratings from readers. This makes it challenging for recognition models to achieve personalization practical scenarios. Existing approaches are generally designed based on the assumption users have consensus whether given...

Feature design is a crucial step in many hyperspectral signal processing applications like signature classification and unmixing, etc. In this paper, we describe technique for automatically designing universal features of signatures. Universality considered both terms the application to multitude problems use specific vs. generic training datasets. The core component our feature non-homogeneous hidden Markov chain (NHMC) characterize wavelet coefficients which capture spectrum semantics...

This paper investigates different machine learning models to solve the historical handwritten manuscript recognition problem. In particular, we test and compare support vector machines, conditional maximum entropy Naive Bayes with kernel density estimates explore their behaviors properties when solving this We focus on a whole word problem avoid having do character segmentation which is difficult degraded documents. Our results publicly available standard dataset of 20 pages George...

Feature selection is a dimensionality reduction technique that selects subset of representative features from high-dimensional data in order to eliminate redundancy. Recently, feature methods based on sparse learning have attracted significant attention due their outstanding performance compared with traditional ignore correlation between features. However, they are restricted by design linear transformations, potential drawback given the underlying structures often non-linear. To leverage...

Hyperspectral signature classification is a quantitative analysis approach for hyperspectral imagery which performs detection and of the constituent materials at pixel level in scene. The procedure can be operated directly on data or performed by using some features extracted from corresponding signatures containing information like signature's energy shape. In this paper, we describe technique that applies non-homogeneous hidden Markov chain (NHMC) models to classification. basic idea use...

We consider the application of non-homogeneous hidden Markov chain (NHMC) models to problem hyperspectral signature classification. It has been previously shown that NHMC model enables detection several semantic structural features signatures. However, there are some aspects spectral data not fully captured by proposed such as relatively smooth but fluctuating regions and fluctuation orientations. In order address these limitations, we propose an improved based on Daubechies-1 wavelets in...

We propose a new spectral unmixing method using semantic representation, which is produced via non-homogeneous hidden Markov chain (NHMC) models applied to wavelet transforms of the spectra. Previous studies have shown that representation robust variability in same materials because it can automatically detect diagnostic features training data. Therefore, our successfully while extracting features, showing high resilience variability. Simulations indicate could be effectively used on Hapke mixtures.