Despite the success of batch normalization (BatchNorm) and a plethora of its variants, the exact reasons for its success are still shady. The original BatchNorm article explained it as a mechanism that reduces the internal covariate shift (ICS), i.e., the distribution shifts in the input of the layers during training. Recently, some articles manifested skepticism on this hypothesis and provided alternative explanations for the success of BatchNorm, such as the applicability of very high learning rates and the ability to smooth the landscape in optimization. In this work, we counter these alternative arguments by demonstrating the importance of reduction in ICS following an empirical approach. We demonstrated various ways to achieve the abovementioned alternative properties without any performance boost. In this light, we explored the importance of different BatchNorm parameters (i.e., batch statistics and affine transformation parameters) by visualizing their effectiveness in the performance and analyzed their connections with ICS. Afterward, we showed a different normalization scheme that fulfills all the alternative explanations except reduction in ICS. Despite having all the alternative properties, we observed its poor performance, which nullifies the alternative claims, rather signifies the importance of the ICS reduction. We performed comprehensive experiments on many variants of BatchNorm, finding that all of them similarly reduce ICS.

Objective: With the scenario of limited labeled dataset, this paper introduces a deep learning-based approach that leverages Diabetic Retinopathy (DR) severity recognition performance using fundus images combined with wide-field swept-source optical coherence tomography angiography (SS-OCTA). Methods: The proposed architecture comprises a backbone convolutional network associated with a Twofold Feature Augmentation mechanism, namely TFA-Net. The former includes multiple convolution blocks extracting representational features at various scales. The latter is constructed in a two-stage manner, i.e., the utilization of weight-sharing convolution kernels and the deployment of a Reverse Cross-Attention (RCA) stream. Results: The proposed model achieves a Quadratic Weighted Kappa rate of 90.2% on the small-sized internal KHUMC dataset. The robustness of the RCA stream is also evaluated by the single-modal Messidor dataset, of which the obtained mean Accuracy (94.8%) and Area Under Receiver Operating Characteristic (99.4%) outperform those of the state-of-the-arts significantly. Conclusion: Utilizing a network strongly regularized at feature space to learn the amalgamation of different modalities is of proven effectiveness. Thanks to the widespread availability of multi-modal retinal imaging for each diabetes patient nowadays, such approach can reduce the heavy reliance on large quantity of labeled visual data. Significance: Our TFA-Net is able to coordinate hybrid information of fundus photos and wide-field SS-OCTA for exhaustively exploiting DR-oriented biomarkers. Moreover, the embedded feature-wise augmentation scheme can enrich generalization ability efficiently despite learning from small-scale labeled data.

The generation of digital maps with lane-level resolution is rapidly becoming a necessity, as semi- or fully-autonomous driving vehicles are now commercially available. In this paper, we present a practical real-time working prototype for road lane detection using LiDAR data, which can be further extended to automatic lane-level map generation. Conventional lane detection methods are limited to simple road conditions and are not suitable for complex urban roads with various road signs on the ground. Given a 3D point cloud scanned by a 3D LiDAR sensor, we categorized the points of the drivable region and distinguished the points of the road signs on the ground. Then, we developed an expectation-maximization method to detect parallel lines and update the 3D line parameters in real time, as the probe vehicle equipped with the LiDAR sensor moved forward. The detected and recorded line parameters were integrated to build a lane-level digital map with the help of a GPS/INS sensor. The proposed system was tested to generate accurate lane-level maps of two complex urban routes. The experimental results showed that the proposed system was fast and practical in terms of effectively detecting road lines and generating lane-level maps.

The extraction and proper utilization of convolutional neural network (CNN) features have a significant impact on the performance of image super-resolution (SR). Although CNN features contain both spatial and channel information, current deep learning techniques for SR often suffer to maximize the performance due to using either the spatial information or channel information. Moreover, they integrate such information within a deep or wide network rather than exploiting all the available features, eventually resulting in high computational complexity. To address these issues, we present a binarized feature fusion (BFF) structure that utilizes the extracted features from global residuals (GR) in an effective way. Each GR consists of multiple hybrid residual attention blocks (HRAB) that effectively integrates the multiscale feature extraction module and channel attention mechanism in a single block. Furthermore, to save computational power, instead of using a large filter size, we use convolutions with different dilation factors to extract multiscale features. We also propose to adopt global skip connections (GSC), short skip connections (SSC), long skip connections (LSC) and GR structure to ease the flow of information without losing important features details. In the paper, we call this overall network architecture as hybrid residual attention network (HRAN). In the experiment, we have observed the efficacy of our method against the state-of-the-art methods for both the quantitative and qualitative comparisons.