An object recognition system has been developed that uses a new class of local image features. The features are invariant to image scaling, translation, and rotation, and partially invariant to illumination changes and affine or 3D projection. These features share similar properties with neurons in inferior temporal cortex that are used for object recognition in primate vision. Features are efficiently detected through a staged filtering approach that identifies stable points in scale space. Image keys are created that allow for local geometric deformations by representing blurred image gradients in multiple orientation planes and at multiple scales. The keys are used as input to a nearest neighbor indexing method that identifies candidate object matches. Final verification of each match is achieved by finding a low residual least squares solution for the unknown model parameters. Experimental results show that robust object recognition can be achieved in cluttered partially occluded images with a computation time of under 2 seconds.

/pdf/object-recognition-from-local-scale-invariant-features-5887b1cgwf.pdf

Object recognition from local scale-invariant features

https://www2.econ.iastate.edu/tesfatsi/DeepLearningInNeuralNetworksOverview.JSchmidhuber2015.pdf

Deep learning in neural networks

http://www.maths.qmul.ac.uk/%7Elatora/report_06.pdf

Complex networks: Structure and dynamics

Functional magnetic resonance imaging (fMRI) is widely used to study the operational organization of the human brain, but the exact relationship between the measured fMRI signal and the underlying neural activity is unclear. Here we present simultaneous intracortical recordings of neural signals and fMRI responses. We compared local field potentials (LFPs), single- and multi-unit spiking activity with highly spatio-temporally resolved blood-oxygen-level-dependent (BOLD) fMRI responses from the visual cortex of monkeys. The largest magnitude changes were observed in LFPs, which at recording sites characterized by transient responses were the only signal that significantly correlated with the haemodynamic response. Linear systems analysis on a trialby-trial basis showed that the impulse response of the neurovascular system is both animal- and site-specific, and that LFPs yield a better estimate of BOLD responses than the multi-unit responses. These findings suggest that the BOLD contrast mechanism reflects the input and intracortical processing of a given area rather than its spiking output.

/pdf/neurophysiological-investigation-of-the-basis-of-the-fmri-1pwsdtcobg.pdf

Neurophysiological investigation of the basis of the fMRI signal

The emergence of a unified cognitive moment relies on the coordination of scattered mosaics of functionally specialized brain regions. Here we review the mechanisms of large-scale integration that counterbalance the distributed anatomical and functional organization of brain activity to enable the emergence of coherent behaviour and cognition. Although the mechanisms involved in large-scale integration are still largely unknown, we argue that the most plausible candidate is the formation of dynamic links mediated by synchrony over multiple frequency bands.

The brainweb: phase synchronization and large-scale integration.

http://brain-mind.med.uoc.gr/sites/default/files/LogothetisCurrBiolITshape1995.pdf

Shape representation in the inferior temporal cortex of monkeys

Functional magnetic resonance imaging (fMRI) has become an essential tool for studying human brain function. Here we describe the application of this technique to anesthetized monkeys. We present spatially resolved functional images of the monkey cortex based on blood oxygenation level dependent (BOLD) contrast. Checkerboard patterns or pictures of primates were used to study stimulus-induced activation of the visual cortex, in a 4.7-Tesla magnetic field, using optimized multi-slice, gradient-recalled, echo-planar imaging (EPI) sequences to image the entire brain. Under our anesthesia protocol, visual stimulation yielded robust, reproducible, focal activation of the lateral geniculate nucleus (LGN), the primary visual area (V1) and a number of extrastriate visual areas, including areas in the superior temporal sulcus. Similar responses were obtained in alert, behaving monkeys performing a discrimination task.

Functional imaging of the monkey brain

A key question concerning the perception of 3D objects is the spatial reference frame used by the brain to represent them. The celerity of the recognition process could be explained by the visual system's ability to quickly transform stored models of familiar 3D objects, or by its ability to specify the relationship among viewpoint-invariant features or volumetric primitives that can be used to accomplish a structural description of an image. Alternatively, viewpoint-invariant recognition could be realized by a system endowed with the ability to perform an interpolation between a set of stored 2D templates, created for each experienced viewpoint. In the present study we set out to examine the nature of object representation in the primate in combined psychophysical-electrophysiological experiments. Monkeys were trained to recognize novel objects from a given viewpoint and subsequently were tested for their ability to generalize recognition for views generated by mathematically rotating the objects around any arbitrary axis. The perception of 3D novel objects was found to be a function of the object's retinal projection at the time of the recognition encounter. Recognition became increasingly difficult for the monkeys as the stimulus was rotated away from its familiar attitude. The generalization field for novel wire-like and spheroidal objects extended to about +/- 40 degrees around an experienced viewpoint. When the animals were trained with as few as three views of the object, 120 degrees apart, they could often interpolate recognition for all views resulting from rotations around the same axis. Recordings from inferotemporal cortex during the psychophysical testing showed a number of neurons with remarkable selectivity for individual views of those objects that the monkey had learned to recognize. Plotting the response of neurons as a function of rotation angle revealed systematic view-tuning curves for rotations in depth. A small percentage of the view-selective cells responded strongly for a particular view and its mirror-symmetrical view. For some of the tested objects, different neurons were found to be tuned to different views of the same object; the peaks of the view-tuning curves were 40-50 degrees apart. Neurons were also found that responded to the sight of unfamiliar objects or distractors. Such cells, however, gave nonspecific responses to a variety of other patterns presented while the monkey performed a simple fixation task.

/pdf/psychophysical-and-physiological-evidence-for-viewer-4dxk1kg0p4.pdf

Psychophysical and Physiological Evidence for Viewer-centered Object Representations in the Primate

nin conjugated to horseradish peroxidase (WGA-HRP) of the eye. After 24 hr, they found signal enhancement to evaluate the specificity of the former by tracing in the optic tract, as well as in the contralateral superior the neuronal connections of the basal ganglia of the colliculus. The time course of the enhanced MRI contrast monkey. Mn 2 and WGA-HRP yielded remarkably simi- indicated that Mn 2 was being transported at approxilar and highly specific projection patterns. By showing mately 2 mm/hr, consistent with fast axonal transport. the sequential transport of Mn 2 from striatum to pal- Intraocular injection of MnCl2 has also been applied in lidum-substantia nigra and then to thalamus, we dem- the rat where labeling was detected in the retina, optic onstrated MRI visualization of transport across at least nerve, chiasm, and contralateral optic tract, as well as one synapse in the CNS of the primate. Transsynaptic in a number of central structures including the dorsal tract tracing in living primates will allow chronic stud- and ventral lateral geniculate nucleus and the superior ies of development and plasticity and provide valuable colliculus (Watanabe et al., 2001). anatomical information for fMRI and electrophysiolog- Axonal transport of radioactive manganese ( 54 Mn 2 ) ical experiments in primates.

J Pauls

Papers

Neurophysiological investigation of the basis of the fMRI signal

Shape representation in the inferior temporal cortex of monkeys

Functional imaging of the monkey brain

Psychophysical and Physiological Evidence for Viewer-centered Object Representations in the Primate

Neurotechnique Magnetic Resonance Imaging of Neuronal Connections in the Macaque Monkey