Translational invariance, or the ability to perceive the same object regardless of its position in the visual field, is enabled by the large receptive fields of inferior temporal neurons.
Because of their convolution and/or pooling operations, convolutional neural networks (CNNs) are assumed to be architecturally invariant to translation. Indeed, several papers have found that these networks are consistently unable to recognize new objects in environments where they have not been trained. Translational invariance, or the ability to perceive the same object regardless of its position in the visual field, is enabled by the large receptive fields of inferior temporal neurons.
Translation invariance means that the system will produce the same result regardless of how the input is translated. A face detector, for example, could report "FACE DETECTED" for all three images in the top row. The property of translational invariance is highly desirable in a variety of tasks, including object recognition and speech recognition.
To know more about temporal neurons refer to: https://brainly.com/question/28495014
#SPJ4
