Contextual associations represented both in neural networks and human behavior

Contextual associations facilitate object recognition in human vision. However, the role of context in artificial vision remains elusive as does the characteristics that humans use to define context. We investigated whether contextually related objects (bicycle-helmet) are represented more similarly in convolutional neural networks (CNNs) used for image understanding than unrelated objects (bicycle-fork). Stimuli were of objects against a white background and consisted of a diverse set of contexts (N=73). CNN representations of contextually related objects were more similar to one another than to unrelated objects across all CNN layers. Critically, the similarity found in CNNs correlated with human behavior across three experiments assessing contextual relatedness, emerging significant only in the later layers. The results demonstrate that context is inherently represented in CNNs as a result of object recognition training, and that the representation in the later layers of the network tap into the contextual regularities that predict human behavior.

ACOUSTIC EXTINGUISHING OF FLAMES DETECTED BY DEEP NEURAL NETWORKS IN EMBEDDED SYSTEMS

This paper presents the application of an intelligent flame and smoke detection platform based on artificial vision using Deep Neural Networks (DNN). An acoustic fire extinguisher can be connected to such a platform. In light of recent research, the use of acoustic technology is an environmentally friendly way of extinguishing flames. One of the advantages is then extinguishing immediately after positive detection of flames or smoke (without unnecessary time delay), which is a new combination in the field of fire protection. The authors contribution is the presentation of the conception of the intelligent acoustic extinguisher and the research work made to accomplish that including algorithms and tests. The main objective of the paper is to present the possibilities of acoustic extinguishing and fire detection using selected deep neural network models in embedded systems.

CFD Modeling and Simulation of an Industrial Hall

The subject of this research is the use of artificial vision techniques to optimize the placement of hot air sources depending on the location of certain equipment in industrial halls operating at low temperatures. Low temperature industrial halls require optimization of the location of hot air sources to achieve low consumption and efficient heating, taking into account the fact that workplace safety standards require the halls are high and at a certain area of windows (to ensure a hall for this reason, it is necessary to optimize the location and hot air sources flow to ensure the heating of these halls.

Dynamical Analysis of a Navigation Algorithm

There is presently a need for more robust navigation algorithms for autonomous industrial vehicles. These have reasonably guaranteed the adequate reliability of the navigation. In the current work, the stability of a modified algorithm for collision-free guiding of this type of vehicle is ensured. A lateral control and a longitudinal control are implemented. To demonstrate their viability, a stability analysis employing the Lyapunov method is carried out. In addition, this mathematical analysis enables the constants of the designed algorithm to be determined. In conjunction with the navigation algorithm, the present work satisfactorily solves the localization problem, also known as simultaneous localization and mapping (SLAM). Simultaneously, a convolutional neural network is managed, which is used to calculate the trajectory to be followed by the AGV, by implementing the artificial vision. The use of neural networks for image processing is considered to constitute the most robust and flexible method for realising a navigation algorithm. In this way, the autonomous vehicle is provided with considerable autonomy. It can be regarded that the designed algorithm is adequate, being able to trace any type of path.

Optimization of the Appearance Quality in CO2 Processed Ready-to-Eat Carrots through Image Analysis

A high-pressure CO2 process applied to ready-to-eat food products guarantees an increase of both their microbial safety and shelf-life. However, the treatment often produces unwanted changes in the visual appearance of products depending on the adopted process conditions. Accordingly, the alteration of the visual appearance influences consumers’ perception and acceptability. This study aims at identifying the optimal treatment conditions in terms of visual appearance by using an artificial vision system. The developed methodology was applied to fresh-cut carrots (Daucus carota) as the test product. The results showed that carrots packaged in 100% CO2 and subsequently treated at 6 MPa and 40 °C for 15 min maintained an appearance similar to the fresh product for up to 7 days of storage at 4 °C. Mild appearance changes were identified at 7 and 14 days of storage in the processed products. Microbiological analysis performed on the optimal treatment condition showed the microbiological stability of the samples up to 14 days of storage at 4 °C. The artificial vision system, successfully applied to the CO2 pasteurization process, can easily be applied to any food process involving changes in the appearance of any food product.

Virtual reality validation of naturalistic modulation strategies to counteract fading in retinal stimulation

Objective: Temporal resolution is a key challenge in artificial vision. Several prosthetic approaches are limited by the perceptual fading of evoked phosphenes upon repeated stimulation from the same electrode. Therefore, implanted patients are forced to perform active scanning, via head movements, to refresh the visual field viewed by the camera. However, active scanning is a draining task, and it is crucial to find compensatory strategies to reduce it. Approach: To address this question, we implemented perceptual fading in simulated prosthetic vision using virtual reality. Then, we quantified the effect of fading on two indicators: the time to complete a reading task and the head rotation during the task. We also tested if stimulation strategies previously proposed to increase the persistence of responses in retinal ganglion cells to electrical stimulation could improve these indicators. Main results: This study shows that stimulation strategies based on interrupted pulse trains and randomisation of the pulse duration allows significant reduction of both the time to complete the task and the head rotation during the task. Significance: The stimulation strategy used in retinal implants is crucial to counteract perceptual fading and to reduce active head scanning during prosthetic vision. In turn, less active scanning might improve the patient's comfort in artificial vision.