Learning invariant object recognition in the visual system with continuous transformations

The cerebral cortex utilizes spatiotemporal continuity in the world to help build invariant representations. In vision, these might be representations of objects. The temporal continuity typical of objects has been used in an associative learning rule with a short-term memory trace to help build invariant object representations. In this paper, we show that spatial continuity can also provide a basis for helping a system to self-organize invariant representations. We introduce a new learning paradigm “continuous transformation learning” which operates by mapping spatially similar input patterns to the same postsynaptic neurons in a competitive learning system. As the inputs move through the space of possible continuous transforms (e.g. translation, rotation, etc.), the active synapses are modified onto the set of postsynaptic neurons. Because other transforms of the same stimulus overlap with previously learned exemplars, a common set of postsynaptic neurons is activated by the new transforms, and learning of the new active inputs onto the same postsynaptic neurons is facilitated. We demonstrate that a hierarchical model of cortical processing in the ventral visual system can be trained with continuous transform learning, and highlight differences in the learning of invariant representations to those achieved by trace learning.     

The effects of continuous and discontinuous groove edges on cell shape and alignment

Nanofabricated model surfaces and digital image analysis of cell shape were used to address the importance of a continuous sharp edge in the alignment of cells to shallow surface grooves. The grooved model surfaces had either continuous or discontinuous edges of various depths (40-400 nm) but identical surface chemistry and groove/ridge dimensions (15 microm wide). Epithelial cells were cultured on the model surfaces for 10 and 24 h. Fluorescence microscopy combined with image analysis were used to quantify cell area and alignment and to make cell shape classifications of individual cells. The degrees of alignment of cells and the percentages of elongated cell classes increased with groove depth on samples with continuous grooves. Two main differences, with regard to cell response, were observed between the continuous and discontinuous grooved surfaces. First, significantly fewer cells aligned to surface grooves with discontinuous edges than to grooves with continuous edges. Second, there were lower percentages of the elongated cell classes on discontinuous grooves than on continuous ones. We concluded that grooved surfaces with continuous edges are more potent in aligning and inducing elongated cells. The results from the present study suggest that a mechanism of alignment involving orientation along a continuous edge is likely.     

Perception of edges and visual texture in the camouflage of the common cuttlefish, Sepia officinalis

The cuttlefish, Sepia officinalis, provides a fascinating opportunity to investigate the mechanisms of camouflage as it rapidly changes its body patterns in response to the visual environment. We investigated how edge information determines camouflage responses through the use of spatially high-pass filtered 'objects' and of isolated edges. We then investigated how the body pattern responds to objects defined by texture (second-order information) compared with those defined by luminance. We found that (i) edge information alone is sufficient to elicit the body pattern known as Disruptive, which is the camouflage response given when a whole object is present, and furthermore, isolated edges cause the same response; and (ii) cuttlefish can distinguish and respond to objects of the same mean luminance as the background. These observations emphasize the importance of discrete objects (bounded by edges) in the cuttlefish's choice of camouflage, and more generally imply that figure-ground segregation by cuttlefish is similar to that in vertebrates, as might be predicted by their need to produce effective camouflage against vertebrate predators.     

Segmentation of moving images by the human visual system

 New segments appearing in an image sequence or spontaneously accelerated segments are band limited by the visual system due to a nonperfect tracking of these segments by eye movements. In spite of this band limitation and acceleration of segments, a coarse segmentation (initial segmentation phase) can be performed by the visual system. This is interesting for the development of purely automatic segmentation algorithms for multimedia applications. In this paper the segmentation of the visual system is modelled and used in an automatic coarse initial segmentation. A suitable model for motion processing based on a spectral representation is presented and applied to the segmentation of synthetic and real image sequences with band limited and accelerated moving foreground and background segments. Received: 1 August 1995/Accepted in revised form: 25 February 1997     

Beyond the edges of a view: boundary extension in human scene-selective visual cortex

To allow perception of a continuous world, cortical mechanisms extrapolate missing information with highly constrained predictions about the environment just beyond the edges of a view. Here, we report functional magnetic resonance imaging evidence for extrapolation of scene layout information beyond what was physically presented, an illusion known as boundary extension. Consistent with behavioral reports, we observed boundary extension for scene-selective attenuation in the parahippocampal place area (PPA) and retrosplenial cortex (RSC), but no such extrapolation of object representations in the lateral occipital complex (LOC). These results demonstrate that scene layout representations are extrapolated beyond the confines of the perceptual input. Such extrapolation may facilitate perception of a continuous world from discontinuous views.     

Apparent position governs contour-element binding by the visual system

An assumption inherent in many models of visual space is that the spatial coordinates of retinal cells implicitly give rise to the perceptual code for position. The results of the experiments reported here, in which it is shown that retinally non-veridical locations of contour elements are used by the visual system for contour-element binding, lend support to a different view. The visual system does not implicitly code position with reference to the labelled locations of retinal cells, but dynamically extracts spatial position from the aggregate result of local computations. These computations may include local spatial relationships between retinal cells, but are not confined to them; other computations, including position derived from local velocity cues, are combined to code the position of objects in the visual world.     

Are direction and speed coded independently by the visual system? Evidence from visual search.

Three visual search experiments examined whether motion is coded as two separate features, speed and direction. Increasing the heterogeneity of the directions in which stimuli moved disrupted detection of a target defined by speed (fast among medium and slow nontargets), suggesting that speed is coded integrally with direction. However, heterogeneity in speed did not disrupt detection of a target moving in a particular direction among nontargets with different directions. This suggests that direction is coded independently of speed. The apparent paradox raised by these contrasting conclusions is consistent with neurophysiological and computational models of motion-detection, which suggest that low-levels of the visual system contain direction-detectors insensitive to speed, while speed is coded at higher levels by detectors which are also sensitive to direction. Evidence consistent with the existence of the latter conjunction detectors was obtained in a final experiment which found search for a conjunction of speed and direction to be parallel.     

Variability of continuous insect cell lines and their identification

Continuous insect cell lines make a special object of research in biology. Insect cells in the established lines differ in the number of attributes from both normal differentiated, and embryonic cells. The period of genome destabilization necessarily precedes cell line immortalization. Genome destabilization is manifested by changes in genome size, cell karyotype, amplification of some retrotransposone families, and induction of their expression. The existence of significant genetic variability in one line puts a problem of searching for invariant attributes providing culture identification and defining the limits of normal polymorphism of cells in the culture. Using the vast collection of insect continuous cell lines stored at the N. I. Vavilov Institute of General Genetics RAS, nine lines were identified by RFLP method of mitochondrial DNA. Variability of DNA-polymorphisms, cellular karyology, morphology, immunological and biochemical attribute in the culture is discussed.     

Motion-based analysis of spatial patterns by the human visual system

BACKGROUND: It is known that the visibility of patterns presented through stationary multiple slits is significantly improved by pattern movements. This study investigated whether this spatiotemporal pattern interpolation is supported by motion mechanisms, as opposed to the general belief that the human visual cortex initially analyses spatial patterns independent of their movements. RESULTS: Psychophysical experiments showed that multislit viewing could not be ascribed to such motion-irrelevant factors as retinal painting by tracking eye movements or an increase in the number of views by pattern movements. Pattern perception was more strongly impaired by the masking noise moving in the same direction than by the noise moving in the opposite direction, which indicates the direction selectivity of the pattern interpolation mechanism. A direction-selective impairment of pattern perception by motion adaptation also indicates the direction selectivity of the interpolation mechanism. Finally, the map of effective spatial frequencies, estimated by a reverse-correlation technique, indicates observers' perception of higher spatial frequencies, the recovery of which is theoretically impossible without the aid of motion information. CONCLUSIONS: These results provide clear evidence against the notion of separate analysis of pattern and motion. The visual system uses motion mechanisms to integrate spatial pattern information along the trajectory of pattern movement in order to obtain clear perception of moving patterns. The pattern integration mechanism is likely to be direction-selective filtering by V1 simple cells, but the integration of the local pattern information into a global figure should be guided by a higher-order motion mechanism such as MT pattern cells.     

Elaboration of lymphotoxin by freshly isolated human T lymphocytes and continuous lymphoid-cell lines.

Freshly isolated human T lymphocytes were demonstrated to produce lymphotoxin (LT) after mitogenic stimulation with phytohemagglutinin (PHA). In contrast, freshly isolated B lymphocytes, stimulated with two B-cell mitogens [pokeweed mitogen (PWM) and Staphylococcus protein A (Staph A)] did not produce the lymphokine, although thymidine incorporation was increased in these cells. We also examined a series of nine continuous human lymphoid-cell lines with B-cell markers and observed the spontaneous release of either large or small amounts of cytotoxin, or none at all. Cytotoxin from one of the productive cell-lines (H4218) was compared in detail with that obtained from PHA-stimulated, freshly isolated human lymphocytes. The behavior of the two cytotoxins was found to be identical in respect to migration on polyacrylamide gel, neutralization with rabbit anti-human α-LT serum, ultracentrifugation on 5–30% sucrose gradients, and stability for 15 min at 75 °C. Observation of these identical parameters strongly suggests that the α-LT elaborated by PHA-stimulated, freshly isolated human lymphoid cells is the same as the cytotoxin obtained from the continuous human lymphoid-cell line H4218. Thus α-LT may also be produced in quantity from continuous lymphoid-cell lines by mass tissue-culture techniques, which are more readily applicable to large-scale production than is purification from freshly cultured human lymphoid tissues. Notably, in cultures of freshly isolated human lymphoid cells, only T cells, and not B cells, generated lymphotoxin. However, continuous human lymphoid-cell lines with B-cell markers can also secrete this lymphokine.     

Processing of complex stimuli and natural scenes in the visual cortex

A major part of vision research builds on the assumption that processing of visual stimuli can be understood on the basis of knowledge about the processing of simplified, artificial stimuli. Recent experimental advances, however, show that a combination of responses to simplified stimuli does not adequately describe responses to natural visual scenes. The systems performance exceeds the performance predicted from understanding its basic constituents. This highlights the fact that the visual system is specifically adapted to the properties of its everyday input and can only fully be understood when probed with naturalistic stimuli.     

Karyological study of the continuous cell lines. Comparative analysis of the Hela and Detroit-6 cell lines]

Comparison of the results of the karyologic analysis of two Hela cell sublines (HeLa1 and HeLa2), obtained from different sources, and of Detroit-6 cell line has shown that all the lines contain marker chromosomes characteristic of the HeLa cell line. Detroit-6 cell line marker chromosomes are similar to markers of the HeLa subline (HeLa1). At the same time, part of marker chromosomes in the two sublines of HeLa cell line (HeLa1 and HeLa2) are different. These data show that HeLa1 and Detroit-6 cell lines are more similar than two sublines of the same HeLa cell line.