Dataset adaptation for the evaluation of small object retrieval in heavily cluttered background

Xerri, Christian, Michael M. Merzenich, Bret E. Peterson, and William Jenkins. Plasticity of primary somatosensory cortex paralleling sensorimotor skill recovery from stroke in adult monkeys. J. Neurophysiol. 79: 2119–2148, 1998. Adult owl and squirrel monkeys were trained to master a small-object retrieval sensorimotor skill. Behavioral observations along with positive changes in the cortical area 3b representations of specific skin surfaces implicated specific glabrous finger inputs as important contributors to skill acquisition. The area 3b zones over which behaviorally important surfaces were represented were destroyed by microlesions, which resulted in a degradation of movements that had been developed in the earlier skill acquisition. Monkeys were then retrained at the same behavioral task. They could initially perform it reasonably well using the stereotyped movements that they had learned in prelesion training, although they acted as if key finger surfaces were insensate. However, monkeys soon initiated alternative strategies for small object retrieval that resulted in a performance drop. Over several- to many-week-long period, monkeys again used the fingers for object retrieval that had been used successfully before the lesion, and reacquired the sensorimotor skill. Detailed maps of the representations of the hands in SI somatosensory cortical fields 3b, 3a, and 1 were derived after postlesion functional recovery. Control maps were derived in the same hemispheres before lesions, and in opposite hemispheres. Among other findings, these studies revealed the following 1) there was a postlesion reemergence of the representation of the fingertips engaged in the behavior in novel locations in area 3b in two of five monkeys and a less substantial change in the representation of the hand in the intact parts of area 3b in three of five monkeys. 2) There was a striking emergence of a new representation of the cutaneous fingertips in area 3a in four of five monkeys, predominantly within zones that had formerly been excited only by proprioceptive inputs. This new cutaneous fingertip representation disproportionately represented behaviorally crucial fingertips. 3) There was an approximately two times enlargement of the representation of the fingers recorded in cortical area 1 in postlesion monkeys. The specific finger surfaces employed in small-object retrieval were differentially enlarged in representation. 4) Multiple-digit receptive fields were recorded at a majority of emergent, cutaneous area 3a sites in all monkeys and at a substantial number of area 1 sites in three of five postlesion monkeys. Such fields were uncommon in area 1 in control maps. 5) Single receptive fields and the component fields of multiple-digit fields in postlesion representations were within normal receptive field size ranges. 6) No significant changes were recorded in the SI hand representations in the opposite (untrained, intact) control hemisphere. These findings are consistent with “substitution” and “vicariation” (adaptive plasticity) models of recovery from brain damage and stroke.

Download Full-text

Learning Multi-view Deep Features for Small Object Retrieval in Surveillance Scenarios

Proceedings of the 23rd ACM international conference on Multimedia - MM '15 ◽

10.1145/2733373.2806349 ◽

2015 ◽

Cited By ~ 6

Author(s):

Haiyun Guo ◽

Jinqiao Wang ◽

Min Xu ◽

Zheng-Jun Zha ◽

Hanqing Lu

Keyword(s):

Small Object ◽

Object Retrieval

Download Full-text

Multi-View Object Detection Based on Deep Learning

Applied Sciences ◽

10.3390/app8091423 ◽

2018 ◽

Vol 8 (9) ◽

pp. 1423 ◽

Cited By ~ 15

Author(s):

Cong Tang ◽

Yongshun Ling ◽

Xing Yang ◽

Wei Jin ◽

Chao Zheng

Keyword(s):

Deep Learning ◽

Object Detection ◽

Regression Models ◽

Detection Methods ◽

Detection Accuracy ◽

Single Shot ◽

Small Object ◽

Object Retrieval ◽

Detection Approach ◽

Small Object Detection

A multi-view object detection approach based on deep learning is proposed in this paper. Classical object detection methods based on regression models are introduced, and the reasons for their weak ability to detect small objects are analyzed. To improve the performance of these methods, a multi-view object detection approach is proposed, and the model structure and working principles of this approach are explained. Additionally, the object retrieval ability and object detection accuracy of both the multi-view methods and the corresponding classical methods are evaluated and compared based on a test on a small object dataset. The experimental results show that in terms of object retrieval capability, Multi-view YOLO (You Only Look Once: Unified, Real-Time Object Detection), Multi-view YOLOv2 (based on an updated version of YOLO), and Multi-view SSD (Single Shot Multibox Detector) achieve AF (average F-measure) scores that are higher than those of their classical counterparts by 0.177, 0.06, and 0.169, respectively. Moreover, in terms of the detection accuracy, when difficult objects are not included, the mAP (mean average precision) scores of the multi-view methods are higher than those of the classical methods by 14.3%, 7.4%, and 13.1%, respectively. Thus, the validity of the approach proposed in this paper has been verified. In addition, compared with state-of-the-art methods based on region proposals, multi-view detection methods are faster while achieving mAPs that are approximately the same in small object detection.

Download Full-text

Revealing gross three-dimensional structure in the SEM

Proceedings, annual meeting, Electron Microscopy Society of America ◽

10.1017/s0424820100127712 ◽

1987 ◽

Vol 45 ◽

pp. 656-657

Author(s):

Sterling P. Newberry

Keyword(s):

Freeze Drying ◽

Three Dimensional ◽

Dimensional Structure ◽

Small Object ◽

Final Solution ◽

Dimensional Representation ◽

X Ray ◽

Three Dimensional Representation ◽

Freeze Dried ◽

Critical Point Drying

The beautiful three dimensional representation of small object surfaces by the SEM leads one to search for ways to open up the sample and look inside. Could this be the answer to a better microscopy for gross biological 3-D structure? We know from X-Ray microscope images that Freeze Drying and Critical Point Drying give promise of adequately preserving gross structure. Can we slice such preparations open for SEM inspection? In general these preparations crush more readily than they slice. Russell and Dagihlian got around the problem by “deembedding” a section before imaging. This some what defeats the advantages of direct dry preparation, thus we are reluctant to accept it as the final solution to our problem. Alternatively, consider fig 1 wherein a freeze dried onion root has a window cut in its surface by a micromanipulator during observation in the SEM.

Download Full-text

Improving x-ray map resolution with image restoration techniques

Proceedings, annual meeting, Electron Microscopy Society of America ◽

10.1017/s0424820100165458 ◽

1996 ◽

Vol 54 ◽

pp. 598-599

Author(s):

Richard B. Mott ◽

John J. Friel ◽

Charles G. Waldman

Keyword(s):

Image Restoration ◽

Hubble Space Telescope ◽

Extensive Literature ◽

Small Object ◽

X Rays ◽

X Ray ◽

Mapping Parameters ◽

The Matrix ◽

The 1960S ◽

Map Resolution

X-rays are emitted from a relatively large volume in bulk samples, limiting the smallest features which are visible in X-ray maps. Beam spreading also hampers attempts to make geometric measurements of features based on their boundaries in X-ray maps. This has prompted recent interest in using low voltages, and consequently mapping L or M lines, in order to minimize the blurring of the maps.An alternative strategy draws on the extensive work in image restoration (deblurring) developed in space science and astronomy since the 1960s. A recent example is the restoration of images from the Hubble Space Telescope prior to its new optics. Extensive literature exists on the theory of image restoration. The simplest case and its correspondence with X-ray mapping parameters is shown in Figures 1 and 2.Using pixels much smaller than the X-ray volume, a small object of differing composition from the matrix generates a broad, low response. This shape corresponds to the point spread function (PSF). The observed X-ray map can be modeled as an “ideal” map, with an X-ray volume of zero, convolved with the PSF. Figure 2a shows the 1-dimensional case of a line profile across a thin layer. Figure 2b shows an idealized noise-free profile which is then convolved with the PSF to give the blurred profile of Figure 2c.

Download Full-text

TILTAI IR KELTAI KAIP XVIII AMŽIAUS VILNIAUS PAVIETO BAJORŲ TEISINIŲ GINČŲ OBJEKTAI IR NUSIKALTIMŲ ERDVĖS

Lietuvos Didžioji Kunigaikštystė Visuomenė. Kasdienybės istorija - XVIII amžiaus studijos / Eighteenth-Century Studies ◽

10.33918/xviiiastudijos/t.4/a14 ◽

2018 ◽

pp. 299-318

Author(s):

Domininkas Burba

Keyword(s):

Eighteenth Century ◽

Rural Areas ◽

Common Type ◽

General Situation ◽

Small Object ◽

Violent Crimes ◽

Rural Space ◽

Green Bridge ◽

Violent Conflicts ◽

The City

Bridges and ferries, as objects of dispute and crime locations among the eighteenth century nobles of Vilnius district, is the main topic of research in this article. Case materials and auxiliary documents from the records of Vilnius district castle and land courts reveal how often bridges are mentioned in the court processes in both violent and non-violent crimes. Research explores what types of violent crimes took place on bridges or ferries most often. It also works on questions of geographic localisation and statistics, discussing general situation of bridges in Vilnius and its neighbouring areas in the eighteenth century. Bridges are regularly mentioned in the books of the eighteenth century Vilnius castle and land courts, albeit most references are not related to conflicts and bridges are mentioned as orientation marks or in reference to location of a real estate object. Both non-violent legal disputes, involving bridges as objects, and violent crimes on the bridges were not in multitude, however non-violent crimes were in smaller numbers. There were seven dispute cases about lands, properties and plots of land where bridges and ferries are mentioned. Non-violent conflicts mostly took place in rural areas of the district, four of them, and three such disputes happened in Vilnius (one on the Green Bridge and two on the bridges over the River Vilnia). Most commonly recorded violent crime on a bridge was beating and, since this was the most common type of crime perpetrated by nobles in the eighteenth century Vilnius district, this trend is logical. A bridge is once mentioned in the record about a raid. In terms of location, more crimes on the bridges took place in the rural space, although this particular space wasn’t dominant, since six crimes were reported in the province and five in the city – two in Vilnius on the Green (Stone) Bridge, two on the bridges over the River Vilnia and one on a ferry near Šnipiškės. Trends in crime locations match general crime tendencies in Vilnius district, where more crimes took place in the rural space than in the urban one. One may guess, that the rare mention of bridges partially testifies to the fact that in the eighteenth century Vilnius district level of communication was not high and there were not too many bridges. On the other hand, when assessing trends in violent crimes in Vilnius district it was revealed that bridge based crimes comprised only one percent of all crimes. Having in mind that bridge is a relatively small object, compared to several different or other urban and rural spaces, this number isn’t that small. Keywords: Vilnius district, castle court, land court, crimes, nobles, peasants, bridges, ferries, passings.

Download Full-text