Detecting 3D lane lines from images is a fundamental challenge and an ill-posed problem in autonomous driving. Existing methods are limited by scene robustness and computational efficiency. This paper introduces an innovative 3D lane detection method that addresses the challenges of lane detection in autonomous driving. Our approach is based on a simple yet efficient view transformation module and layer-by-layer refined bird’s-eye-view (BEV) features. First, we introduce a module for dual-branch view transformation that combines deformable convolutions and view relation modules to convert front view features into BEV features. This enhances scene robustness across various data scenarios. Additionally, we suggest an auxiliary training head for inverse view transformation that offers supplementary supervisory information. Moreover, we progressively refine the BEV features, making use of features from different levels. The results of our experiment indicate the supremacy of our approach on two datasets, as it achieved a considerable increase in F1-score in comparison to preexisting methods.

In this paper, we present TrimTail, a simple but effective emission regularization method to improve the latency of streaming ASR models. The core idea of TrimTail is to apply length penalty (i.e., by trimming trailing frames, see Fig. 1-(b)) directly on the spectrogram of input utterances, which does not require any alignment. We demonstrate that TrimTail is computationally cheap and can be applied online and optimized with any training loss or any model architecture on any dataset without any extra effort by applying it on various end-to-end streaming ASR networks either trained with CTC loss [1] or Transducer loss [2]. We achieve 100 ~ 200ms latency reduction with equal or even better accuracy on both Aishell-1 and Librispeech. Moreover, by using TrimTail, we can achieve a 400ms algorithmic improvement of User Sensitive Delay (USD) with an accuracy loss of less than 0.2.

Federated learning faces huge challenges from model overfitting due to the lack of data and statistical diversity among clients. To address these challenges, this paper proposes a novel personalized federated learning method via Bayesian variational inference named pFedBayes. To alleviate the overfitting, weight uncertainty is introduced to neural networks for clients and the server. To achieve personalization, each client updates its local distribution parameters by balancing its construction error over private data and its KL divergence with global distribution from the server. Theoretical analysis gives an upper bound of averaged generalization error and illustrates that the convergence rate of the generalization error is minimax optimal up to a logarithmic factor. Experiments show that the proposed method outperforms other advanced personalized methods on personalized models, e.g., pFedBayes respectively outperforms other SOTA algorithms by 1.25%, 0.42% and 11.71% on MNIST, FMNIST and CIFAR-10 under non-i.i.d. limited data.

Extracting robust feature representation is one of the key challenges in person/vehicle re-identification (ReID). Although approaches based on convolutional neural networks (CNN) for local feature learning have been very successful in person ReID, they have not been discussed deeply about effective ways to obtain local features for slicing. The horizontal slicing is commonly used for person ReID, which is more likely to cause fine-grained information to be overlooked in vehicle ReID with diverse orientations and a discriminative information distribution different from that of the person. To overcome these limitations, we propose the stronger collaborative attention network (SCAN). Specifically, we first perform cross slicing of backbone features in each scale branches to ensure that different fine-grained information can be divided into different parts with higher probability, and fuse different directional features of the same order of parts in each local scale branch to gradually strengthen the global nature of local vehicle features by associating other regions with common regions of different directions. To further mine the useful information, we design three modules carefully. (i) The collaborative feature refinement module (CFR) is proposed to further explore the collaborative relationship of local features and filter out the redundancy of collaborative information in local scale branches by superposition fusion strategy. (ii) The discriminative edge information learning module (DEL) is proposed to learn reliable discriminative edge information by adding edge parts in each local scale branch. (iii) The nonvisual information embedding module (NIE) is introduced to mitigate the feature bias caused by background and vehicle shape by embedding learnable explicit nonvisual camera and view information. Experimental results of our proposed method are superior, which achieve state-of-the-art performance on vehicle ReID benchmarks.

Video-based person re-identification (Re-ID) methods can extract richer features than image-based ones from short video clips. The existing methods usually apply simple strategies, such as average/max pooling, to obtain the tracklet-level features, which has been proved hard to aggregate the information from all video frames. In this paper, we propose a simple yet effective Temporal Aggregation with Clip-level Attention Network (TACAN) to solve the temporal aggregation problem in a hierarchal way. Specifically, a tracklet is firstly broken into different numbers of clips, through a two-stage temporal aggregation network we can get the tracklet-level feature representation. A novel min-max loss is introduced to learn both a clip-level attention extractor and a clip-level feature representer in the training process. Afterwards, the resulting clip-level weights are further taken to average the clip-level features, which can generate a robust tracklet-level feature representation at the testing stage. Experimental results on four benchmark datasets, including the MARS, iLIDS-VID, PRID-2011 and DukeMTMC-VideoReID, show that our TACAN has achieved significant improvements as compared with the state-of-the-art approaches.

By dealing with robust modeling and online learning together in a unified random vector functional-link networks (RVFLNs) framework, this paper presents a novel robust online sequential RVFLNs for data modeling of dynamic time-varying systems together with its application for a blast furnace (BF) ironmaking process. First, to overcome the difficulties caused by the nonlinear time-varying dynamics of process and to enable the RVFLNs to learn online and to avoid data saturation, an improved online sequential version of RVFLNs (OS-RVFLNs) is presented by sequential learning with forgetting factor. It has been shown that the improved OS-RVFLNs with forgetting factor is not only suitable for the large-scale and real-time data transfer situation but also can adjust the sensitivity of the algorithm to different samples. Second, in order to solve the issue of modeling robustness when the dataset is contaminated with various outliers, a Cauchy distribution function weighted M-estimator is introduced to strengthen the robustness of the improved OS-RVFLNs. The non-Gaussian Cauchy distribution function is used to estimate the weights of different data and thus the corresponding contribution on modeling can be properly distinguished. Experiments using actual industrial data of a large BF ironmaking process have demonstrated that the proposed algorithm produces a much stronger robustness and better estimation accuracy than other algorithms.

How to measure and evaluate software quality is one important research subject in the computer software field. For helping people understand, measure and evaluate software quality, International Organization for Standardization (ISO), has been carrying out years of researches on software quality standardization. The thoughts and methods shown in standardization will help people further understand and predict software quality easier. This paper introduces the development process of the international quality standards and gives more focus on the software quality development trends.

Deep learning models (DLMs) are state-of-the-art techniques in speech recognition.However, training good DLMs can be time consuming especially for production-size models and corpora.Although several parallel training algorithms have been proposed to improve training efficiency, there is no clear guidance on which one to choose for the task in hand due to lack of systematic and fair comparison among them.In this paper we aim at filling this gap by comparing four popular parallel training algorithms in speech recognition, namely asynchronous stochastic gradient descent (ASGD), blockwise model-update filtering (BMUF), bulk synchronous parallel (BSP) and elastic averaging stochastic gradient descent (EASGD), on 1000-hour LibriSpeech corpora using feed-forward deep neural networks (DNNs) and convolutional, long short-term memory, DNNs (CLDNNs).Based on our experiments, we recommend using BMUF as the top choice to train acoustic models since it is most stable, scales well with number of GPUs, can achieve reproducible results, and in many cases even outperforms single-GPU SGD.ASGD can be used as a substitute in some cases.

In cognitive radio (CR) networks, cross-layer design is an important issue since the behavior of one protocol could affect the performance of others. However, for the energy-constraint CR networks, the previous works mostly focus on maximizing the throughput in physical or transport layer, rather than the energy efficiency of the end-to-end transmission control protocol (TCP). In this paper, we propose a novel cross-layer scheme which takes the lower layers' parameters into consideration, e.g., signal-to-noise ratio (SNR), modulation and frame size, to improve the energy efficiency of TCP. Specifically, we use a finite state Markov channel (FSMC) model to characterize the fading channel, and solve the optimization problem by a restless bandit approach. Simulation results show that the physical and data link layer parameters affect the energy efficiency of the TCP traffic significantly and the performance can be improved by the adjustment of lower layers' parameters compared with the existing method.