Biases in object location estimation: the role of camera rotations and translation

The current study investigated a systematic bias in spatial memory in which people, following a perspective shift from encoding to recall, indicated the location of an object further to the direction of the shit. In Experiment 1, we documented this bias by asking participants to encode the position of an object in a virtual room and then indicate it from memory following a perspective shift induced by camera translation and rotation. In Experiment 2, we decoupled the influence of camera translations and camera rotations and examined also whether adding more information in the scene would reduce the bias. We also investigated the presence of age-related differences in the precision of object location estimates and the tendency to display the bias related to perspective shift. Overall, our results showed that camera translations led to greater systematic bias than camera rotations. Furthermore, the use of additional spatial information improved the precision with which object locations were estimated and reduced the bias associated with camera translation. Finally, we found that although older adults were as precise as younger participants when estimating object locations, they benefited less from additional spatial information and their responses were more biased in the direction of camera translations. We propose that accurate representation of camera translations requires more demanding mental computations than camera rotations, leading to greater uncertainty about the position of an object in memory. This uncertainty causes people to rely on an egocentric anchor thereby giving rise to the systematic bias in the direction of camera translation.

Correcting Imprecise Object Locations for Training Object Detectors in Remote Sensing Applications

Object detection on aerial and satellite imagery is an important tool for image analysis in remote sensing and has many areas of application. As modern object detectors require accurate annotations for training, manual and labor-intensive labeling is necessary. In situations where GPS coordinates for the objects of interest are already available, there is potential to avoid the cumbersome annotation process. Unfortunately, GPS coordinates are often not well-aligned with georectified imagery. These spatial errors can be seen as noise regarding the object locations, which may critically harm the training of object detectors and, ultimately, limit their practical applicability. To overcome this issue, we propose a co-correction technique that allows us to robustly train a neural network with noisy object locations and to transform them toward the true locations. When applied as a preprocessing step on noisy annotations, our method greatly improves the performance of existing object detectors. Our method is applicable in scenarios where the images are only annotated with points roughly indicating object locations, instead of entire bounding boxes providing precise information on the object locations and extents. We test our method on three datasets and achieve a substantial improvement (e.g., 29.6% mAP on the COWC dataset) over existing methods for noise-robust object detection.

Comparable performance on a spatial memory task in data collected in the lab and online

Online data collection offers a wide range of benefits including access to larger and more diverse populations, together with a reduction in the experiment cycle. Here we compare performance in a spatial memory task, in which participants had to estimate object locations following viewpoint shifts, using data from a controlled lab-based setting and from an unsupervised online sample. We found that the data collected in a conventional laboratory setting and those collected online produced very similar results, although the online data was more variable with standard errors being about 10% larger than those of the data collected in the lab. Overall, our findings suggest that spatial memory studies using static images can be successfully carried out online with unsupervised samples. However, given the higher variability of the online data, it is recommended that the online sample size is increased to achieve similar standard errors to those obtained in the lab. For the current study and data processing procedures, this would require an online sample 25% larger than the lab sample.

Flexible Use of Spatial Frames of Reference for Object–Location Memory in Older Adults

In memory, representations of spatial features are stored in different reference frames; features relative to our position are stored egocentrically and features relative to each other are stored allocentrically. Accessing these representations engages many cognitive and neural resources, and so is susceptible to age-related breakdown. Yet, recent findings on the heterogeneity of cognitive function and spatial ability in healthy older adults suggest that aging may not uniformly impact the flexible use of spatial representations. These factors have yet to be explored in a precisely controlled task that explicitly manipulates spatial frames of reference across learning and retrieval. We used a lab-based virtual reality task to investigate the relationship between object–location memory across frames of reference, cognitive status, and self-reported spatial ability. Memory error was measured using Euclidean distance from studied object locations to participants’ responses at testing. Older adults recalled object locations less accurately when they switched between frames of reference from learning to testing, compared with when they remained in the same frame of reference. They also showed an allocentric learning advantage, producing less error when switching from an allocentric to an egocentric frame of reference, compared with the reverse direction of switching. Higher MoCA scores and better self-assessed spatial ability predicted less memory error, especially when learning occurred egocentrically. We suggest that egocentric learning deficits are driven by difficulty in binding multiple viewpoints into a coherent representation. Finally, we highlight the heterogeneity of spatial memory performance in healthy older adults as a potential cognitive marker for neurodegeneration, beyond normal aging.

Generating accurate 3D gaze vectors using synchronized eye tracking and motion capture

Assessing gaze behaviour during real-world tasks is difficult; dynamic bodies moving through dynamic worlds make finding gaze fixations challenging. Current approaches involve laborious coding of pupil positions overlaid on video. One solution is to combine eye tracking with motion tracking to generate 3D gaze vectors. When combined with tracked or known object locations, fixation detection can be automated. Here we use combined eye and motion tracking and explore how linear regression models generate accurate 3D gaze vectors. We compare spatial accuracy of models derived from four short calibration routines across three data types: the performance of calibration routines were assessed using calibration data, a validation task that demands short fixations on task-relevant locations, and an object interaction task we used to bridge the gap between laboratory and "in the wild" studies. Further, we generated and compared models using spherical and cartesian coordinate systems and monocular (Left or Right) or binocular data. Our results suggest that all calibration routines perform similarly, with the best performance (i.e., sub-centimeter errors) coming from the task (i.e., the most "natural") trials when the participant is looking at an object in front of them. Further, we found that spherical coordinate systems generate more accurate gaze vectors with no differences in accuracy when using monocular or binocular data. Overall, we recommend recording one-minute calibration datasets, using a binocular eye tracking headset (for redundancy), a spherical coordinate system when depth is not considered, and ensuring data quality (i.e., tracker positioning) is high when recording datasets.

The Black Box effect: sensory stimulation after learning interferes with the retention of long-term object location memory in rats

Reducing sensory experiences during the period that immediately follows learning improves long-term memory retention in healthy humans, and even preserves memory in patients with amnesia. To date, it is entirely unclear why this is the case, and identifying the neurobiological mechanisms underpinning this effect requires suitable animal models, which are currently lacking. Here, we describe a straightforward experimental procedure in rats that future studies can use to directly address this issue. Using this method, we replicated the central findings on quiet wakefulness obtained in humans: We show that rats that spent 1 h alone in a familiar dark and quiet chamber (the Black Box) after exploring two objects in an open field expressed long-term memory for the object locations 6 h later, while rats that instead directly went back into their home cage with their cage mates did not. We discovered that both visual stimulation and being together with conspecifics contributed to the memory loss in the home cage, as exposing rats either to light or to a cage mate in the Black Box was sufficient to disrupt memory for object locations. Our results suggest that in both rats and humans, everyday sensory experiences that normally follow learning in natural settings can interfere with processes that promote long-term memory retention, thereby causing forgetting in form of retroactive interference. The processes involved in this effect are not sleep-dependent because we prevented sleep in periods of reduced sensory experience. Our findings, which also have implications for research practices, describe a potentially useful method to study the neurobiological mechanisms that might explain why normal sensory processing after learning impairs memory both in healthy humans and in patients suffering from amnesia.

Feature Extraction Model with Group-Based Classifier for Content Extraction from Video Data

In today's PC illustration, numerous object locations of videos are quite critical duties to accomplish. Swiftly and reliably recognising and distinguishing the multiple aspects of a video is a crucial attribute for collaborating with one's condition (object). The core issue is that in theory, to ensure that no significant aspect is missing; all aspects of a content in a video must be scanned for elements on various different scales. It requires some investment and effort anyway, to really arrange the substance of a given content region and both time and computational limits that an operator can spend on classification are constrained. Two presumption procedures for accelerating the standard identifier are performed by the proposed method and demonstrate their capability by performing both identification efficiency and velocity. The main enhancement of our group-based classifier focuses on accelerating the grouping of sub features by planning the problem as a selection procedure for consecutive features. The subsequent improvement gives better multiscale features to distinguish objects of all sizes without rescaling the information image from a video. Extracting contents from video is an assortment of successive images with a steady time interim. So video can give more data about contents in it when situations are changing regarding time. Along these lines, physically taking care of contents with features are very unimaginable. In the proposed work, it is suggested that a Group-based Video Content Extraction Classifier (GbCCE) extracts content from a video by extracting relevant features using a group-based classifier. The proposed method is distinct from conventional approaches and the findings indicate that better output is demonstrated by the proposed method.

ViTT: Vision Transformer Tracker

This paper presents a new model for multi-object tracking (MOT) with a transformer. MOT is a spatiotemporal correlation task among interest objects and one of the crucial technologies of multi-unmanned aerial vehicles (Multi-UAV). The transformer is a self-attentional codec architecture that has been successfully used in natural language processing and is emerging in computer vision. This study proposes the Vision Transformer Tracker (ViTT), which uses a transformer encoder as the backbone and takes images directly as input. Compared with convolution networks, it can model global context at every encoder layer from the beginning, which addresses the challenges of occlusion and complex scenarios. The model simultaneously outputs object locations and corresponding appearance embeddings in a shared network through multi-task learning. Our work demonstrates the superiority and effectiveness of transformer-based networks in complex computer vision tasks and paves the way for applying the pure transformer in MOT. We evaluated the proposed model on the MOT16 dataset, achieving 65.7% MOTA, and obtained a competitive result compared with other typical multi-object trackers.

An Object Detection Framework Based on Deep Features and High-Quality Object Locations

Objection detection has long been a fundamental issue in computer vision. Despite being widely studied, it remains a challenging task in the current body of knowledge. Many researchers are eager to develop a more robust and efficient mechanism for object detection. In the extant literature, promising results are achieved by many novel approaches of object detection and classification. However, there is ample room to further enhance the detection efficiency. Therefore, this paper proposes an image object detection and classification, using a deep neural network (DNN) for based on high-quality object locations. The proposed method firstly derives high-quality class-independent object proposals (locations) through computing multiple hierarchical segments with super pixels. Next, the proposals were ranked by region score, i.e., several contours wholly enclosed in the proposed region. After that, the top-ranking object proposal was adopted for post-classification by the DNN. During the post-classification, the network extracts the eigenvectors from the proposals, and then maps the features with the softmax classifier, thereby determining the class of each object. The proposed method was found superior to traditional approaches through an evaluation on Pascal VOC 2007 Dataset.