Search Results (5)

Detecting vulnerable road users is a major challenge for autonomous vehicles due to their small size. Various sensor modalities have been investigated, including mono or stereo cameras and 3D LiDAR sensors, which are limited by environmental conditions and hardware costs. Radar sensors are a low-cost and robust option, with high-resolution 4D radar sensors being suitable for advanced detection tasks. However, they involve challenges such as few and irregularly distributed measurement points and disturbing artifacts. Learning-based approaches utilizing pillar-based networks show potential in overcoming these challenges. However, the severe sparsity of radar data makes detecting small objects with only a few points difficult. We extend a pillar network with our novel Sparsity-Robust Feature Fusion (SRFF) neck, which combines high- and low-level multi-resolution features through a lightweight attention mechanism. While low-level features aid in better localization, high-level features allow for better classification. As sparse input data are propagated through a network, the increasing effective receptive field leads to feature maps of different sparsities. The combination of features with different sparsities improves the robustness of the network for classes with few points. Full article

Federated learning (FL) is a field in distributed optimization. Therein, the collection of data and training of neural networks (NN) are decentralized, meaning that these tasks are carried out across multiple clients with limited communication and computation capabilities. In FL, the client NNs are first trained with locally available data. Next, they are aggregated to update a global NN. FL suffers from non-independent and identically distributed (iid) data and asynchronous communication between the server and the clients, which degrades the NN’s overall performance. In this work, we investigate FL for a small-live-gesture-sensing NN, using a low-power 60 GHz frequency modulated continuous wave radar from Infineon Technologies. The challenges of data sparsity, i.e., only a fraction of a gesture recording corresponds to an executed gesture combined with non-iid data, pose issues during neural network training. It is shown that FL reaches an accuracy higher than 96.2% for an iid setting. However, an increasing level of non-iid data degrades the accuracy to 64.8%. To tackle the accuracy degradation, we propose to dynamically adapt the class weights during the training procedure based on each client’s varying ratio of data sparsity. Moreover, regularization terms are included in the loss function to prevent client drift and overconfidence in the client’s NN prediction. Finally, it is shown that the proposed modifications increase the NN’s performance, such that an accuracy of 97% is obtained despite a high degree of non-iid data. Full article

Deep neural network-based autoencoders can effectively extract high-level abstract features with outstanding generalization performance but suffer from sparsity of extracted features, insufficient robustness, greedy training of each layer, and a lack of global optimization. In this study, the broad learning system (BLS) is improved to obtain a new model for data reconstruction. Support Vector Domain Description (SVDD) is one of the best-known one-class-classification methods used to solve problems where the proportion of sample categories of data is extremely unbalanced. The SVDD is sensitive to penalty parameters C, which represents the trade-off between sphere volume and the number of target data outside the sphere. The training process only considers normal samples, which leads to a low recall rate and weak generalization performance. To address these issues, we propose a BLS-based weighted SVDD algorithm (BLSW_SVDD), which introduces reconstruction error weights and a small number of anomalous samples when training the SVDD model, thus improving the robustness of the model. To evaluate the performance of BLSW_SVDD model, comparison experiments were conducted on the UCI dataset, and the experimental results showed that in terms of accuracy and F1 values, the algorithm has better performance advantages than the traditional and improved SVDD algorithms. Full article

Image set matching (ISM) has attracted increasing attention in the field of computer vision and pattern recognition. Some studies attempt to model query and gallery sets under a joint or collaborative representation framework, achieving impressive performance. However, existing models consider only the competition and collaboration among gallery sets, neglecting the inter-instance relationships within the query set which are also regarded as one important clue for ISM. In this paper, inter-instance relationships within the query set are explored for robust image set matching. Specifically, we propose to represent the query set instances jointly via a combined dictionary learned from the gallery sets. To explore the commonality and variations within the query set simultaneously to benefit the matching, both low rank and class-level sparsity constraints are imposed on the representation coefficients. Then, to deal with nonlinear data in real scenarios, the‘kernelized version is also proposed. Moreover, to tackle the gross corruptions mixed in the query set, the proposed model is extended for robust ISM. The optimization problems are solved efficiently by employing singular value thresholding and block soft thresholding operators in an alternating direction manner. Experiments on five public datasets demonstrate the effectiveness of the proposed method, comparing favorably with state-of-the-art methods. Full article

As a widely used classifier, sparse representation classification (SRC) has shown its good performance for hyperspectral image classification. Recent works have highlighted that it is the collaborative representation mechanism under SRC that makes SRC a highly effective technique for classification purposes. If the dimensionality and the discrimination capacity of a test pixel is high, other norms (e.g., ${\ell }_2$ -norm) can be used to regularize the coding coefficients, except for the sparsity ${\ell }_1$ -norm. In this paper, we show that in the kernel space the nonnegative constraint can also play the same role, and thus suggest the investigation of kernel fully constrained least squares (KFCLS) for hyperspectral image classification. Furthermore, in order to improve the classification performance of KFCLS by incorporating spatial-spectral information, we investigate two kinds of spatial-spectral methods using two regularization strategies: (1) the coefficient-level regularization strategy, and (2) the class-level regularization strategy. Experimental results conducted on four real hyperspectral images demonstrate the effectiveness of the proposed KFCLS, and show which way to incorporate spatial-spectral information efficiently in the regularization framework. Full article