scholarly journals Differentiable Patch Selection for Image Recognition

Author(s):  
Jean-Baptiste Cordonnier ◽  
Aravindh Mahendran ◽  
Alexey Dosovitskiy ◽  
Dirk Weissenborn ◽  
Jakob Uszkoreit ◽  
...  
1970 ◽  
Vol 48 (4) ◽  
pp. 717-740 ◽  
Author(s):  
Edward Mitchell

Dolphins and porpoises have complex pigmentation patterns. A terminology is devised for the complicated pattern on Delphinus delphis, and is applied to related species. A Stenella–Delphinus species group is recognized, with species which can be segregated into four pigmentation pattern types: STRIPED, CRISSCROSS, SADDLED, and SPOTTED. Morphological derivations indicate that saddled is like simple countershading; could independently have given rise to striped, spotted, or crisscross patterns; and is thus inferred to be the most generalized pattern of the four. Crisscross pattern is the most complex, possibly the most specialized, and occurs only on Delphinus. Other morphological evidence from rostral grooves and cranial air sinuses likewise indicates that Delphinus is more specialized than Stenella. The saddled pattern on Stenella is thus generalized and probably also primitive within the Delphinidae.An overlook of other delphinid species confirms that the saddled pattern is generalized and probably primitive. The Lissodelphinae show a generalized pattern; that of the Orcininae can be related to it. Species of the genus Lagenorhynchus show a gradation in the extent and increasing complexity of the dorsal flank blaze from L. obscurus and L. australis through L. obliquidens, L. albirostris, and L. cruciger to L. acutus. The two allopatric species L. cruciger and L. acutus are both specialized in the subdivision of the flank patch. Selection for blazes dorsoanteriorly on the trunk resulted in extension of flank blazes in Lagenorhynchus, of thoracic blazes in Stenella. L. acutus and D. delphis, among the most specialized representatives of genera Lagenorhynchus and Delphinus, respectively, independently evolved a yellow-brown side patch—it is on the flank of the former, on the thorax of the latter. The Cephalorhynchinae show patterns related to saddled, and perhaps paralleled the Orcininae in pigmentation trends. Cephalorhynchus heavisidei appears to be a mimic of Orcinus orca. C. commersoni and the allopatric phocoenid Phocoenoides truei appear to be convergent—both are disruptively pigmented in the same way. Additional pigmentation types recognized are UNIFORM, LOBED, and DISRUPTIVE.In the Stenella–Delphinus species group: the saddled pattern is possibly for obliterative shading through counterlighting for concealment; the striped pattern is possibly for disguise through disruptive pigmentation and camouflage; the spotted pattern on a countershaded base is possibly for camouflage by resemblance to background; and the crisscross pattern is possibly for disguise through disruptive pigmentation, and camouflage by countershading and shadow mimicry. On most delphinids, eye patches, bridle stripes, and lip patches serve to mask eyes, blowhole, and mouth. The thoracic patch may be a false shadow to conceal young swimming in echelon formation.Striped or Euphrosyne dolphins of Pacific and Atlantic waters are conspecific. Distinct subspecies of Stenella caeruleoalba are disallowed as yet unproved.


2011 ◽  
Vol 480-481 ◽  
pp. 323-328
Author(s):  
Yong Mei Zhang ◽  
Li Ma

The realization of object recognition is discussed, the intrinsic relation between factors through high step association analysis is presented to mine the various features using property related analysis, the sensitive recognition features are chosen to improve the recognition efficiency. A multi-agent object recognition model(MAORM) is established, which combines concurrency research results and the specific characteristics of multi-sensor remote sensing image recognition. A majority-decision algorithm based on multi-agent is proposed. Experiment results show the system can effectively identify the bridges, wharfs, ships and so on. Compared with a single remote sensing image, the system can distinguish targets with higher recognition.


2021 ◽  
pp. 821-827
Author(s):  
Antonio J. Tallón-Ballesteros ◽  
Luís Correia ◽  
Rocío Leal-Díaz

Author(s):  
Yanfeng Lu ◽  
Lihao Jia ◽  
Hong Qiao ◽  
Yi Li ◽  
Zongshuai Qi

Biologically inspired model (BIM) for image recognition is a robust computational architecture, which has attracted widespread attention. BIM can be described as a four-layer structure based on the mechanisms of the visual cortex. Although the performance of BIM for image recognition is robust, it takes the randomly selected ways for the patch selection, which is sightless, and results in heavy computing burden. To address this issue, we propose a novel patch selection method with oriented Gaussian–Hermite moment (PSGHM), and we enhanced the BIM based on the proposed PSGHM, named as PBIM. In contrast to the conventional BIM which adopts the random method to select patches within the feature representation layers processed by multi-scale Gabor filter banks, the proposed PBIM takes the PSGHM way to extract a small number of representation features while offering promising distinctiveness. To show the effectiveness of the proposed PBIM, experimental studies on object categorization are conducted on the CalTech05, TU Darmstadt (TUD) and GRAZ01 databases. Experimental results demonstrate that the performance of PBIM is a significant improvement on that of the conventional BIM.


2020 ◽  
Vol 98 (5) ◽  
pp. 331-341
Author(s):  
E.P. McNeill ◽  
I.D. Thompson ◽  
P.A. Wiebe ◽  
G.M. Street ◽  
J. Shuter ◽  
...  

Multi-scale selection patterns can be understood from two perspectives: coarse-scale patterns as the summation of fine-scale patterns (scaling-up), or as a hierarchy produced from multiple contributory factors with differential effects on organismal fitness (hierarchical). We examined woodland caribou (Rangifer tarandus caribou (Gmelin, 1788)) selection of foraging locations across two spatiotemporal scales to test whether selection patterns between them were consistent (scaling-up) or different (hierarchical) to determine which framework most accurately describes their foraging behaviour. Seven adult female woodland caribou were equipped with GPS telemetry radio collars outfitted with high-definition video cameras that recorded woodland caribou foraging choices throughout the summer. Fine-scale data from videos combined with direct measurements in the field along movement trajectories obtained from GPS fixes were used to estimate (i) feeding station selection and (ii) food patch selection. We estimated resource selection functions for each scale following a use–availability structure. Woodland caribou exhibited resource selection at both scales. Apart from selection for species of the lichen Cladina (Nyl.) Nyl. and patches associated with high abundance of Cladina, few patterns were consistent across both scales. Our study suggests that even at very fine scales, woodland caribou selection for foraging locations is hierarchical in nature.


2019 ◽  
Vol 91 ◽  
pp. 103798
Author(s):  
Bolei Xu ◽  
Jingxin Liu ◽  
Xianxu Hou ◽  
Bozhi Liu ◽  
Guoping Qiu

Sign in / Sign up

Export Citation Format

Share Document