More precisely, our model includes feature detectors at each retinal location. When a visual feature appears at a given location, a memory box opens and processes information about the corresponding visual feature in a leaky integrator21. Once stimulation at this retinal location terminates, the ...
Using psychophysical procedures, the edge-elements allow characterisation of the response properties of these feature-detectors in detail. At the same time, the (task-irrelevant) high-level representation of the stimulus, within which the edge probe is embedded, can be manipulated independently by ...
The pipeline of our method is shown in Fig.2, with example results displayed in Fig.1and the algorithm listed in Algorithm 1. In the first step, the GPU-SLIC method [1,34] is exploited to over-segment an input image into superpixels, which serve as seed regions in the\(1^{st}\)le...
To prevent moving target problems, CCN trains the feature detectors one by one to obtain the best possible detection from them. Although the initial hidden-neuron weights are static, once they have been trained, the neurons are not touched again, so the features they identify are permanently pro...
[12]. Two-stage detectors learn separate models for object region detection and object recognition. A typical example is the Region-based Convolutional Neural Network (R-CNN) [13]. Advanced models such as Fast R-CNN [14] and Faster R-CNN [15] have also been developed. In particular, ...