Search

The most recent 3D object detectors for point clouds rely on the coarse voxel-based representation rather than the accurate point-based representation due to a higher box recall in the voxel-based Region Proposal Network (RPN). However, the detection accuracy is severely restricted by the information loss of pose details in the voxels. Different from considering the point cloud as voxel or point representation only, we propose a point-to-voxel feature learning approach to voxelize the point cloud with both the point-wise semantic and local spatial features, which maintains the voxel-wise features to build the high-recall voxel-based RPN and also provides the accurate point-wise features for refining the detection results. Another difficulty in object detection for point cloud is that the visible part varies a lot against the full view of object because of the perspective issues in data acquisition. To address this, we propose an attentive corner aggregation module to attentively aggregate the features of local point cloud surrounding a 3D proposal from the perspectives of eight corners in the proposal 3D bounding box. The experimental results on the competitive KITTI 3D object detection benchmark show that the proposed method achieves state-of-the-art performance.

The main challenge for scene parsing arises when complex scenes with highly diverse objects are encountered. The objects not only differ in scale and appearance but also in semantics. Previous works focus on encoding the multi-scale contextual information (via pooling or atrous convolutions) generally on top of compact high-level features (i.e., at a single stage). In this work, we argue that a rich set of cues exist at multiple stages of the network, encapsulating low, mid and high-level scene details. Therefore, an optimal scene parsing model must aggregate multi-scale context at all three levels of the feature hierarchy; a capability that lacks in state-of-the-art scene parsing models. To address this limitation, we introduce a novel architecture with three new blocks that systematically aggregate low, mid and high tier features. The heart of our approach is a high-level feature aggregation module that augments sparsely connected atrous convolution with dense local and layer-wise connections to avoid gridding artifacts. Besides, we employ a novel feature pyramid augmentation and semantic refinement unit to generate low- and mid-level features that are mixed with high-level features at the decoder. We extensively evaluate our proposed approach on the large-scale Cityscapes and ADE2K benchmarks. Our approach surpasses many latest models on both datasets, achieving mean intersection-over-union (mIoU) scores of 80.5% and 44.0% on Cityscapes and ADE20K, respectively.

Recently, ${L_{1}}$ -norm-based robust discriminant feature extraction technique has been attracted much attention in dimensionality reduction and pattern recognition. However, it does not relate to the scatter matrix which well characterizes the geometric structure of data. In this paper, we propose a robust formulation of graph embedding framework for dimensionality reduction. In this robust framework, we use ${L_{2}}$ -norm to measure the distance along space dimension and ${L_{1}}$ -norm to sum overall data points. The proposed robust graph embedding framework retains the traditional framework’s desirable properties, such as rotational invariance and well geometric structure, and simultaneously suppresses outliers. Based on this framework, we develop a simple and robust feature extraction method, namely ${L_{2,1}}$ -norm-based discriminant locality preserving projections ( ${L_{2,1}}$ -DLPP) and provide an effective iterative algorithm to solve ${L_{2,1}}$ -DLPP. Extensive experiments in artificial data and three popular face databases illustrate the effectiveness of our proposed method.

Diabetic retinopathy (DR) is one of the most common microvascular complications and its early detection is critical for the prevention of vision loss. Recent studies have indicated that microaneurysms (MAs) are the hallmark of DR. However, the detection of MAs relies on trained clinicians and relatively expensive software. Moreover, manual errors often lower the accuracy of this detection. Therefore, an automatic analysis technique is highly demanded in the detection of DR progression. In this paper, we present a novel and complete methodology involving two different ways from the view of MAs turnover and pathological risk factors to diagnose the progression of DR. Specifically, one approach follows the traditional image analysis-based roadmap to obtain MAs turnover. The other investigates seven pathological features, related with MAs turnover, to classify the unchanged, new, and resolved MAs by means of statistical analysis and pattern classification techniques. The evaluations on Grampian diabetes database show that the proposed image analysis method could achieve a sensitivity of 94% and a specificity of 93%, while the classification model could achieve 89% sensitivity and 88% specificity, respectively. We also analyzed the potential weight of pathological risk factors leading to the MAs turnover, which could provide an alternative guidance for the progression of DR than traditional detection methods. In conclusion, this study provides a novel and noninvasive detection technique for early diagnosis of diabetic retinopathy with a competitive accuracy.