This paper presents a new deep-learning based method to simultaneously calibrate the intrinsic parameters of fisheye lens and rectify the distorted images. Assuming that the distorted lines generated by fisheye projection should be straight after rectification, we propose a novel deep neural network to impose explicit geometry constraints onto processes of the fisheye lens calibration and the distorted image rectification. In addition, considering the nonlinearity of distortion distribution in fisheye images, the proposed network fully exploits multi-scale perception to equalize the rectification effects on the whole image. To train and evaluate the proposed model, we also create a new largescale dataset labeled with corresponding distortion parameters and well-annotated distorted lines. Compared with the state-of-the-art methods, our model achieves the best published rectification quality and the most accurate estimation of distortion parameters on a large set of synthetic and real fisheye images.
We present evidence for parallel auditory thalamocortical pathways that serve two different behaviors. The pallid bat listens for prey-generated noise (5-35 kHz) to localize prey, while reserving echolocation [downward frequency-modulated (FM) sweeps, 60-30 kHz] for obstacle avoidance. Its auditory cortex contains a tonotopic map representing frequencies from 6 to 70 kHz. The high-frequency (BF > 30 kHz) representation is dominated by FM sweep-selective neurons, whereas most neurons tuned to lower frequencies prefer noise. Retrograde tracer injections into these physiologically distinct cortical regions revealed that the high-frequency region receives input from the suprageniculate (SG) nucleus, but not the ventral division of the medial geniculate body (MGBv), in all experiments (n = 9). In contrast, the low-frequency region receives tonotopically organized input from the MGBv in all experiments (n = 16). Labeling in the SG was observed in only two of these experiments. Both cortical regions also receive sparse inputs from medial (MGBm) and parts of the dorsal division (MGBd) outside the SG. These results show that the low- and high-frequency regions of a single tonotopic map receive dominant inputs from different thalamic divisions. Within the low-frequency region, most neurons are binaurally inhibited, and an orderly map of interaural intensity difference (IID) sensitivity is present. We show that the input to the IID map arises from topographically organized projections from the MGBv. As observed in other species, a frequency-dependent organization is observed in the lateromedial direction in the MGBv. These data demonstrate that MGBv-to-auditory cortex connections are organized with respect to both frequency and binaural selectivity.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.