Innate Preferences Affect Results of Object Recognition Task in Wild Type and Alzheimer’s Disease Mouse Models

Background: Object recognition task (ORT) is a widely used behavioral paradigm to assess memory in rodent models, due to its easy technical execution, the lack of aversive stressful stimuli, and the possibility to repeat the test on the same animals. However, mouse exploration might be strongly influenced by a variety of variables. Objective: To study whether innate preferences influenced exploration in male and female wild type mice and the Alzheimer’s disease (AD) model 3xTg. Methods: We first evaluated how object characteristics (material, size, and shape) influence exploration levels, latency, and exploration modality. Based on these findings, we evaluated whether these innate preferences biased the results of ORT performed in wild type mice and AD models. Results: Assessment of Exploration levels, i.e., the time spent in exploring a certain object in respect to the total exploration time, revealed an innate preference for objects made in shiny materials, such as metal and glass. A preference for bigger objects characterized by higher affordance was also evident, especially in male mice. When performing ORT, exploration was highly influenced by these innate preferences. Indeed, both wild type and AD mice spent more time in exploring the metal object, regardless of its novelty. Furthermore, the use of objects with higher affordance such as the cube was a confounding factor leading to “false” results that distorted ORT interpretation. Conclusion: When designing exploration-based behavioral experiments aimed at assessing memory in healthy and AD mice, object characteristics should be carefully evaluated to improve scientific outcomes and minimize possible biases.

Generating Images with Physics-Based Rendering for an Industrial Object Detection Task: Realism versus Domain Randomization

Limited training data is one of the biggest challenges in the industrial application of deep learning. Generating synthetic training images is a promising solution in computer vision; however, minimizing the domain gap between synthetic and real-world images remains a problem. Therefore, based on a real-world application, we explored the generation of images with physics-based rendering for an industrial object detection task. Setting up the render engine’s environment requires a lot of choices and parameters. One fundamental question is whether to apply the concept of domain randomization or use domain knowledge to try and achieve photorealism. To answer this question, we compared different strategies for setting up lighting, background, object texture, additional foreground objects and bounding box computation in a data-centric approach. We compared the resulting average precision from generated images with different levels of realism and variability. In conclusion, we found that domain randomization is a viable strategy for the detection of industrial objects. However, domain knowledge can be used for object-related aspects to improve detection performance. Based on our results, we provide guidelines and an open-source tool for the generation of synthetic images for new industrial applications.

Equivalent noise characterization of human lightness constancy

A goal of visual perception is to provide stable representations of task-relevant scene properties (e.g. object reflectance) despite variation in task-irrelevant scene properties (e.g. illumination, reflectance of other nearby objects). To study such representational stability in the context of lightness representations in humans, we introduce a threshold-based psychophysical paradigm. We measure how thresholds for discriminating the achromatic reflectance of a target object (task-relevant property) in rendered naturalistic scenes are impacted by variation in the reflectance functions of background objects (task-irrelevant property). We refer to these thresholds as lightness discrimination thresholds. Our approach has roots in the equivalent noise paradigm. This paradigm relates signals to internal and external sources of noise and has been traditionally used to investigate contrast coding. For low variation in background reflectance, lightness discrimination thresholds were nearly constant, indicating that observers' internal noise determines threshold in this regime. As background object reflectance variation increases, its effects start to dominate performance. We report lightness discrimination thresholds as a function of the amount of variability in the background object reflectance to determine the equivalent noise - the smallest level of task-irrelevant (i.e. background reflectance) variation that substantially corrupts the visual representation (i.e. perceived object lightness) of the task-relevant variable (i.e. achromatic reflectance). A linear receptive field model, which employs a single center-surround receptive field tailored to our stimulus set, captures human behavior in this task. Our approach provides a method for characterizing the effect of task-irrelevant scene variations on the perceptual representation of a task-relevant scene property.

"Gravitationally Lensed Gravitation" (GLG)

The intention of this paper is to point out a remarkable hitherto unknown effect of General Relativity. Starting from fundamental physical principles and phenomena arising from General Relativity, it is demonstrated by a simple Gedankenexperiment that a gravitational lens enhances not only the light intensity of a background object but also its gravitational field strength by the same factor. Thus, multiple images generated by a gravitational lens are not just optical illusions, they also have a gravitational effect at the location of the observer! The "Gravitationally Lensed Gravitation" (GLG) may help to better understand the rotation curves of galaxies since it leads to an enhancement of the gravitational interactions of the stars. Furthermore, it is revealed that besides a redshift of the light of far distant objects, the cosmic expansion also causes a corresponding weakening of their gravitational effects. The explanations are presented entirely without metric representation and tensor formalism. Instead, the behavior of light is used to indicate the effect of spacetime curvature. The gravitation is described by the field strength which is identical to the free fall acceleration. The new results thus obtained provide a reference for future numerical calculations based on the Einstein field equations.

An improved object tracking algorithm based on adaptive weighted strategy and occlusion detection mechanism

Due to complex background and volatile object shape-appearance in image, the stability and accuracy of tracking algorithm is often disturbed and reduced. So how to accurately and robustly track object in object tracking application is a challenge topic at home and abroad. Built upon the methodologies of compressive tracking and spatio-temporal context, a simple yet robust object tracking method is proposed for solving the drift and occlusion problems in paper. It combines two existing classical ideas into a single framework: adaptive weighted idea and occlusion detection mechanism. In order to weaken interference problems of object background, object area is firstly partitioned into equal-sized sub-patches and the different weight related with location information is assigned for each patch; Then, for improving its robustness, Bhattacharyya distance is adopted to find out these samples with maximum discrimination; In addition, our proposed occlusion detection mechanism is for recapturing the tracked object when occlusion occurs. Many simulation experiments show that our proposed algorithm achieves more favorable performance than these existing state-of-the-art algorithms in handing various challenging infrared videos, especially occlusion and shape deformation.

ISPY – NaCo Imaging Survey for Planets around Young stars

Context. Within the NaCo-ISPY exoplanet imaging program, we aim at detecting and characterizing the population of low-mass companions at wide separations (≳10 AU), focusing in particular on young stars either hosting a known protoplanetary disk or a debris disk. Aims. R CrA is one of the youngest (1–3 Myr) and most promising objects in our sample because of two previous studies that suggested the presence of a close companion. Our aim is to directly image and characterize the companion for the first time. Methods. We observed R CrA twice with the NaCo instrument at the Very Large Telescope (VLT) in the L′ filter with a one year time baseline in between. The high-contrast imaging data were reduced and analyzed and the companion candidate was detected in both datasets. We used artificial negative signals to determine the position and brightness of the companion and the related uncertainties. Results. The companion is detected at a separation of 196.8 ± 4.5/196.6 ± 5.9 mas (18.7 ± 1.3/18.7 ± 1.4 AU) and position angle of 134.7 ± 0.5 ° /133.7 ± 0.7° in the first/second epoch observation. We measure a contrast of 7.29 ± 0.18/6.70 ± 0.15 mag with respect to the primary. A study of the stellar proper motion rejects the hypothesis that the signal is a background object. The companion candidate orbits in the clockwise direction and, if on a face-on circular orbit, its period is ∼43 − 47 yr. This value disagrees with the estimated orbital motion and therefore a face-on circular orbit may be excluded. Depending on the assumed age, extinction, and brightness of the primary, the stellar companion has a mass between 0.10 ± 0.02 M⊙ and 1.03−0.18+0.20 M⊙ range, if no contribution from circumsecondary material is taken into account. Conclusions. As already hypothesized by previous studies, we directly detected a low-mass stellar companion orbiting the young Herbig Ae/Be star R CrA. Depending on the age assumptions, the companion is among the youngest forming companions imaged to date, and its presence needs to be taken into account when analyzing the complex circumstellar environment of R CrA.

Embedded Deep Learning for Ship Detection and Recognition

Ship detection and recognition are important for smart monitoring of ships in order to manage port resources effectively. However, this is challenging due to complex ship profiles, ship background, object occlusion, variations of weather and light conditions, and other issues. It is also expensive to transmit monitoring video in a whole, especially if the port is not in a rural area. In this paper, we propose an on-site processing approach, which is called Embedded Ship Detection and Recognition using Deep Learning (ESDR-DL). In ESDR-DL, the video stream is processed using embedded devices, and we design a two-stage neural network named DCNet, which is composed of a DNet for ship detection and a CNet for ship recognition, running on embedded devices. We have extensively evaluated ESDR-DL, including performance of accuracy and efficiency. The ESDR-DL is deployed at the Dongying port of China, which has been running for over a year and demonstrates that it can work reliably for practical usage.