Search

Extracting informative local features in point clouds is crucial for accurately understanding spatial information inside 3D point data. Previous works utilize either complex network designs or simple multi-layer perceptrons (MLP) to extract the local features. However, complex networks often incur high computational cost, whereas simple MLP may struggle to capture the spatial relations among local points effectively. These challenges limit their scalability to delicate and real-time tasks, such as autonomous driving and robot navigation. To address these challenges, we propose a novel 3D Directional Encoding Network (3D-DENet) capable of effectively encoding spatial relations with low computational cost. 3D-DENet extracts spatial and point features separately. The key component of 3D-DENet for spatial feature extraction is Directional Encoding (DE), which encodes the cosine similarity between direction vectors of local points and trainable direction vectors. To extract point features, we also propose Local Point Feature Multi-Aggregation (LPFMA), which integrates various aspects of local point features using diverse aggregation functions. By leveraging DE and LPFMA in a hierarchical structure, 3D-DENet efficiently captures both detailed spatial and high-level semantic features from point clouds. Experiments show that 3D-DENet is effective and efficient in classification and segmentation tasks. In particular, 3D-DENet achieves an overall accuracy of 90.7% and a mean accuracy of 90.1% on ScanObjectNN, outperforming the current state-of-the-art method while using only 47% floating point operations.

Semantic segmentation has been a crucial technology for various practical applications such as autonomous driving. Recently, attempts have been made to improve the performance of semantic segmentation using depth information. However, most attempts have been focused on the indoor environment for the following reasons. First, it is relatively more difficult to obtain accurate and dense depth information outdoors. Second, a network with a new structure is required to use depth information because processing depth as an input demands an additional encoder. To overcome aforementioned difficulties, we propose a novel Depth and Pixel-distance based Attention (DPA) module, which utilizes depth information to compute the similarity between pixels. The similarity of pixels is computed using the fact that pixels belonging to the same object have similar depth values. Because only the relative difference in depth is considered, it is relatively robust despite the accuracy of the provided depth information. Furthermore, DPA is a simple plug-in module that can be applied to existing RGB-based segmentation backbones. Since no encoder is added, it is much more efficient in terms of computation. We conduct extensive experiments on the Cityscapes dataset using various baseline architectures. Regardless of the baseline models, DPA yields meaningful performance improvements in semantic segmentation tasks. It is also computationally more efficient compared to the methods that take depth information as input.

Downsampling layers are essential for convolutional neural network-based semantic segmentation methods to widen their receptive fields. However, as fine-grained information is lost in the layers, the accuracy of these methods becomes limited. The need for downsampling layers can be eliminated by using a transformer encoder. Nevertheless, removing downsampling layers inevitably increases the computational cost of the network. In this paper, we present a mask transformer layer that reduces computational cost in any transformer-based networks by substituting a vanilla transformer layer. Additionally, we introduce an aggregation scheme to merge masked outputs, which enhances the accuracy of predictions. Our method aggregates intermediate outputs to generate a final output where the number of intermediate outputs depends on the importance of an area. With this strategy, we achieve different computational cost levels by modulating the threshold used to determine the importance. Our method comprises the following steps. First, we split the transformer encoder into several blocks and attach a segmentation decoder to each block to estimate the intermediate segmentation output. On the basis of the intermediate outputs and predefined thresholds, we classify unnecessary image patches and remove them in subsequent blocks. By progressively masking unnecessary patches, we obtain multiple intermediate outputs for important areas; aggregating them yields better segmentation accuracy with a lower computational burden. In addition, we determine the most effective training scheme and devise a threshold-search algorithm to optimally determine threshold hyperparameters. Extensive experiments on the ADE20K, Cityscapes, and Pascal-Context datasets verify the efficacy of our design, which surpasses the accuracy of the baseline method with lower computational cost.

Traditionally, the success of the Transformer has been attributed to its token mixer, particularly the self-attention mechanism. However, recent studies suggest that replacing such attention-based token mixer with alternative techniques can yield comparable results in various vision tasks. This highlights the importance of the model’s overall architectural structure in achieving optimal performance, rather than relying exclusively on the specific choice of the token mixer. Building on this insight, we introduce Discrete Wavelet TransFormer, an innovative framework that incorporates Discrete Wavelet Transform to elevate all the building blocks of the Transformer to a higher standard. By exploiting distinct attributes of Discrete Wavelet Transform, Discrete Wavelet TransFormer not only strengthens the network’s ability to learn more intricate feature representations across different levels of abstraction, but also facilitates lossless down-sampling to promote a more resilient and efficient network. A comprehensive evaluation is conducted on diverse vision tasks, and the results conclusively demonstrate that Discrete Wavelet TransFormer outperforms all other state-of-the-art Transformer-based models across all tasks by a significant margin.

Abstract Accurate real‐time localisation is crucial for various autonomous driving applications. Numerous methods for localization, combining various kinds of input such as data from environmental sensors, inertial measurement units, and the Global Positioning System, have been proposed. Given the present state of affairs with regards to map compactness and sensor affordability, lane map‐matching‐based methods with cameras used as environmental sensors have been widely proposed. In previous methods, however, lanes detected by cameras could falsely correspond to other lanes in a map. This would result in inaccuracies such as a two‐lane or three‐lane deviation from the actual pose. This paper proposes a localisation method integrating the ego‐lane index to avoid lane deviations during lane map matching. A novel ego‐lane index identification algorithm and methodology to integrate the ego‐lane index information into the lane‐based map‐matching function is presented. It is demonstrated that incorporating ego‐lane index information improves the accuracy of map‐matching localisation. The efficacy of our approach in diverse environments, including complex urban scenarios, was successfully demonstrated.

Abstract Ni-rich layered oxides are promising cathode materials due to their high capacities. However, their synthesis process retains a large amount of Li residue on the surface, which is a main source of gas generation during operation of the battery. In this study, combined with simulation and experiment, we propose the optimal metal phosphate coating materials for removing residual Li from the surface of the Ni-rich layered oxide cathode material LiNi0.91Co0.06Mn0.03O2. First-principles-based screening process for 16 metal phosphates is performed to identify an ideal coating material that is highly reactive to Li2O. By constructing the phase diagram, we obtain the equilibrium phases from the reaction of coating materials and Li2O, based on a database using a DFT hybrid functional. Experimental verification for this approach is accomplished with Mn3(PO4)2, Co3(PO4)2, Fe3(PO4)2, and TiPO4. The Li-removing capabilities of these materials are comparable to the calculated results. In addition, electrochemical performances up to 50 charge/discharge cycles show that Mn-, Co-, Fe-phosphate materials are superior to an uncoated sample in terms of preventing capacity fading behavior, while TiPO4 shows poor initial capacity and rapid reduction of capacity during cycling. Finally, Li-containing equilibrium phases examined from XRD analysis are in agreement with the simulation results.