Search

This article focuses on the first aspect of the album of deep learning: the deep convolutional method. The traditional matching point extraction algorithm typically uses manually designed feature descriptors and the shortest distance between them to match as the matching criterion. The matching result can easily fall into a local extreme value, which causes missing of the partial matching point. Targeting this problem, we introduce a two-channel deep convolutional neural network based on spatial scale convolution, which performs matching pattern learning between images to realize satellite image matching based on a deep convolutional neural network. The experimental results show that the method can extract the richer matching points in the case of heterogeneous, multi-temporal and multi-resolution satellite images, compared with the traditional matching method. In addition, the accuracy of the final matching results can be maintained at above 90%.

Numerous deep learning-based object detection methods have achieved excellent performance. However, the performance on small-size object detection and positive and negative sample imbalance problems is not satisfactory. We propose a multi-scale feature selective matching network (MFSMNet) to improve the performance of small-size object detection and alleviate the positive and negative sample imbalance problems. First, we construct a multi-scale semantic enhancement module (MSEM) to compensate for the information loss of small-sized targets during down-sampling by obtaining richer semantic information from features at multiple scales. Then, we design the anchor selective matching (ASM) strategy to alleviate the training dominated by negative samples caused by the imbalance of positive and negative samples, which converts the offset values of the localization branch output in the detection head into localization scores and reduces negative samples by discarding low-quality anchors. Finally, a series of quantitative and qualitative experiments on the Microsoft COCO 2017 and PASCAL VOC 2007 + 2012 datasets show that our method is competitive compared to nine other representative methods. MFSMNet runs on a GeForce RTX 3090. [ABSTRACT FROM AUTHOR]

Based on the detection technology, regressing predicted bounding boxes provides an effective approach in multiple object tracking. However, if only the information in the current frame is considered, identity (ID) switch is easy to happen when objects interact. In this paper, we propose an Object-Aware Bounding Box Regression (OABBR) for online multi-object tracking. We first propose an Object-Aware Spatial-Temporal Understanding (OASTU) module to mine the correlated information in corresponding object's trajectory. OASTU updates features of predictions by the correlated information. By using the updated features, we further perform bounding box regression. Besides, to make features extracted by the backbone network contain more ID information, we construct a weak ID constraint in the training phase. The introduced weak ID constraint facilitates OASTU to be ID consistent and further alleviates ID switch. By exploring the spatial-temporal information in corresponding object's trajectory, each prediction is able to know the information of the corresponding object, which makes the purpose of its regression clearer. Experimental results on four public benchmarks demonstrate the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

Cartoonization as a special type of artistic style transfer is a difficult image processing task. The current existing artistic style transfer methods cannot generate satisfactory cartoon-style images due to that artistic style images often have delicate strokes and rich hierarchical color changes while cartoon-style images have smooth surfaces without obvious color changes, and sharp edges. To this end, we propose a cartoon loss based generative adversarial network (CartoonLossGAN) for cartoonization. Particularly, we first reuse the encoder part of the discriminator to build a compact generative adversarial network (GAN) based cartoonization architecture. Then we propose a novel cartoon loss function for the architecture. It can imitate the process of sketching to learn the smooth surface of the cartoon image, and imitate the coloring process to learn the coloring of the cartoon image. Furthermore, we also propose an initialization strategy, which is used in the scenario of reusing the discriminator to make our model training easier and more stable. Extensive experimental results demonstrate that our proposed CartoonLossGAN can generate fantastic cartoon-style images, and outperforms four representative methods. [ABSTRACT FROM AUTHOR]

Appearance similarity is of great importance for the association between objects and candidates. Recurrent models and similarity vector are two ways widely used by trackers for calculating similarities between objects and candidates. Recurrent models, like Long Short Term Memory network (LSTM), are capable of modeling the continuous change of object's appearance in trajectory. But it is prone to identity (ID) switch when only employ recurrent models as appearance model. The similarity vector way is able to maintain correct IDs for objects when they reappear. But association fails easily when the object is partially occluded and similarity vector is used as the only appearance model. To obtain more accurate and robust appearance similarity, in this paper, we propose an online association by continuous-discrete appearance similarity measurement, OA-CDASM, for multi-object tracking. For continuous perspective, the concept of "smoothness" is proposed to explicitly model and use the continuous and smooth change of object's appearance in trajectory. For discrete perspective, similarity vector is employed. By taking both continuous smoothness and discrete similarity vector into consideration, we can get the continuous-discrete appearance similarity measurement, CDASM, and further perform online association based on CDASM. Experimental results on three public benchmarks demonstrate the effectiveness of our work. [ABSTRACT FROM AUTHOR]

Spectral analysis was usually used to guide unsupervised feature selection. However, the performances of these methods are not always satisfactory due to that they may generate continuous pseudo labels to approximate the discrete real labels. In this paper, a novel unsupervised feature selection method is proposed based on self-expression model. Unlike existing spectral analysis based methods, we utilize self-expression model to capture the relationships between the features without learning the cluster labels. Specifically, each feature can be reconstructed by using a linear combination of all the features in the original feature space, and a representative feature should give a large weight to reconstruct other features. Besides, a structure preserved constraint is incorporated into our model for keeping the local manifold structure of the data. Then an efficient alternative iterative algorithm is utilized to solve our proposed model with the theoretical analysis on its convergence. The experimental results on different datasets show the effectiveness of our method. [ABSTRACT FROM AUTHOR]