This study proposes a generic approach which performs a series of systematic analyses by first introducing a data volume decomposition method to generate useful data features for performing semantic segmentation analy...
详细信息
This study proposes a generic approach which performs a series of systematic analyses by first introducing a data volume decomposition method to generate useful data features for performing semantic segmentation analysis involving 3d point-cloud data. Pipeline parallelism protocol is then implemented to accelerate the deep learning model's training phase. Our proposed approach is verified by decomposing around 2.0 billion point-clouddatapoints, as extracted from an open-source Semantic3ddataset, into many 3d regular structures with defined numbers of voxels. Each derived3d structure has imposed normality in their datadistribution of the respective label classes. Using the optimal hyperparameters for model training, the resulting trained model achieves average overall accuracy (mOA) and average intersection over union (mIOU) values of 0.984 and 0.752, respectively, on a testing dataset having close to 800 million point-clouddatapoints. The results are comparable with that of other state-of-the-art models in the literature.
Implementing cross-modal hashing between 2d images and3d point-cloud data is a growing concern in real-world retrieval systems. Simply applying existing cross-modal approaches to this new task fails to adequately cap...
详细信息
ISBN:
(纸本)9798350390155;9798350390162
Implementing cross-modal hashing between 2d images and3d point-cloud data is a growing concern in real-world retrieval systems. Simply applying existing cross-modal approaches to this new task fails to adequately capture latent multi-modal semantics and effectively bridge the modality gap between 2d and3d. To address these issues without relying on hand-crafted labels, we propose contrastive masked autoencoders based self-supervised hashing (CMAH) for retrieval between images andpoint-clouddata. We start by contrasting 2d-3d pairs and explicitly constraining them into a joint Hamming space. This contrastive learning process ensures robust discriminability for the generated hash codes and effectively reduces the modality gap. Moreover, we utilize multi-modal auto-encoders to enhance the model's understanding of multi-modal semantics. By completing the masked image/point-clouddata modeling task, the model is encouraged to capture more localized clues. In addition, the proposed multi-modal fusion block facilitates fine-grained interactions among different modalities. Extensive experiments on three public datasets demonstrate that the proposed CMAH significantly outperforms all baseline methods.
To overcome the limitations of pavement skid resistance prediction using the friction coefficient, a genetic-algorithm-improved neural network (GAI-NN) was developed in this study. First, three-dimensional (3d) point-...
详细信息
To overcome the limitations of pavement skid resistance prediction using the friction coefficient, a genetic-algorithm-improved neural network (GAI-NN) was developed in this study. First, three-dimensional (3d) point-clouddata of an asphalt pavement surface were obtained using a smart sensor (Gocator 3110). The friction coefficient of the pavement was then obtained using a pendulum friction tester. The 3d point-cloud dataset was then analyzed to recover the missing data and perform denoising. In particular, these data were filled using cubic spline interpolation. Parameters for texture characterization were defined, and methods for computing the parameters were developed. Finally, the GAI-NN model was developed by modifying the weights and thresholds. The test results indicated that using pavement surface texture 3ddata, the GAI-NN was capable of predicting the pavement friction coefficient with sufficient accuracy, with an error of 12.1%.
暂无评论