Objective. Computer vision-based assistive technology solutions can revolutionise the quality of care for people with sensorimotor disorders. The goal this work was to enable trans-radial amputees use a simple, yet efficient, computer vision system grasp and move common household objects two-channel myoelectric prosthetic hand. Approach. We developed deep learning-based artificial augment functionality commercial prosthesis. Our main conceptual novelty is that we classify regards pattern...

Changes in pig behaviours are a useful aid detecting early signs of compromised health and welfare. In commercial settings, automatic detection through visual imaging remains challenge due to farm demanding conditions, e.g., occlusion one from another. Here, two deep learning-based detector methods were developed identify postures drinking group-housed pigs. We first tested the system ability detect changes these measures at group-level during routine management. then demonstrated our...

Automated, vision-based early warning systems have been developed to detect behavioural changes in groups of pigs monitor their health and welfare status. In commercial settings, automatic recording feeding behaviour remains a challenge due problems variation illumination, occlusions similar appearance different pigs. Additionally, such systems, which rely on pig tracking, often overestimate the actual time spent feeding, inability identify and/or exclude non-nutritive visits (NNV) area. To...

Change in the frequency of contact between pigs within a group may be indicative change physiological or health status one more group, occurrence abnormal behaviour, e.g. tail-biting. Here, we developed novel framework that detects and quantifies interaction, i.e., pig head to another rear, groups. The method does not require individual tracking/identification uses only inexpensive camera-based data capturing infrastructure. We modified architecture well-established deep learning models...

The human visual cortex has evolved to determine efficiently objects from within a scene. Hierarchical MAX (HMAX) is an object recognition model which been inspired by the cortex, and sparse coding, characteristic of neurons in was previously integrated into HMAX for improved performance. In this study, order further enhance accuracy, we have developed elastic net-regularized dictionary learning approach use model. We term En-HMAX With model, can exploit sparsity-grouping tradeoff, such that...

Dictionary learning was introduced for sparse image representation. Today, it is a cornerstone of classification. We propose novel dictionary method to recognise images handwritten numbers. Our focus maximise the sparse-representation and discrimination power class-specific dictionaries. We, first time, adopt new feature space, i.e., histogram oriented gradients (HOG), generate columns (atoms). The HOG features robustly describe fine details hand-writings. design an objective function...

We enhance the efficacy of an existing dictionary pair learning algorithm by adding a incoherence penalty term. After presenting alternating minimization solution, we apply proposed incoherent (InDPL) method in classification novel open-source database Chinese numbers. Benchmarking results confirm that InDPL offers enhanced accuracy, especially when number training samples is limited.

We modified an automated method capable of quantifying behaviors which we then applied to the changes associated with post-weaning transition in pigs. The is data-driven and depends solely on video-captured image data without relying sensors or additional pig markings. It was video images generated from experiment during post-weaned piglets were subjected treatments either containing not in-feed antimicrobials (ZnO antibiotics). These expected affect piglet performance health short-term by...

Abstract The assessment of livestock welfare aids in keeping an eye on the health, physiology, and environment animals order to prevent deterioration, detect injuries, stress, sustain productivity. Because it puts more consumer pressure farming industries change how are treated make them humane, has also grown be a significant marketing tactic. Common visual procedures followed by experts vets could expensive, subjective, need specialized staff. Recent developments artificial intelligence...

Visual processing has attracted a lot of attention in the last decade. Hierarchical approaches for object recognition are gradually becoming widely-accepted. Generally, they inspired by ventral stream human visual cortex, which is charge rapid categorization. Similar to objects, natural scenes share common features and can, therefore, be classified same manner. However, generally show high level statistical correlation between classes. This, fact, major challenge most models. Rapid...

<div>Quantitative trading through automated systems has been vastly growing in recent years. The advancement machine learning algorithms pushed that growth even further, where their capability extracting high-level patterns within financial markets data is evident. Nonetheless, with supervised can be challenging since the system learns to predict price minimize error rather than optimize a performance measure. Reinforcement Learning (RL), paradigm intersects optimal control theory,...

Deep learning methods have become the key ingredient in field of computer vision; particular, convolutional neural networks (CNNs). Appropriating network architecture and data pre-processing significant impact on performance. This paper focuses classification handwritten Chinese numbers. Firstly, we applied various to our collected image dataset. Secondly, customised a CNN-based with minimal number layers parameters specifically for task. Experimental results showed that proposed provides...

<div>Quantitative trading through automated systems has been vastly growing in recent years. The advancement machine learning algorithms pushed that growth even further, where their capability extracting high-level patterns within financial markets data is evident. Nonetheless, with supervised can be challenging since the system learns to predict price minimize error rather than optimize a performance measure. Reinforcement Learning (RL), paradigm intersects optimal control theory,...

Humans can recognise objects under partial occlusion. Machine-based approaches cannot reliably and scenes in the presence of This paper investigates use elastic net hierarchical MAX (En-HMAX) model to handle occlusions. Our experiments show that En-HMAX achieves an accuracy ~70%, when ~50% artificial occlusions are applied centre visual object-field. Furthermore, same percentage occlusion is peripheral, reports higher accuracies. A similar degree robustness has been observed recognising...

<p>We propose Conditional Value-at-Risk (CVaR) investment agents to solve the problems of single asset trading and assets allocation under Direct Reinforcement Learning framework. We two convex CVaR-based agents, CVaR-constrained unconstrained CVaR-sensitive. Convexity allows conveniently implementing incremental learning, leading an adaptive investing agent at efficient computational cost with merit guaranteed policy convergence. Our experiments frictional various markets reveal...