In the field of object detection, deep learning models have achieved great success in recent years. Despite these advances, detecting small objects remains difficult. Most objects in aerial images have features that are a challenge for traditional object detection techniques, including small size, high density, high variability, and varying orientation. Previous approaches have used slicing methods on high-resolution images or feature maps to improve performance. However, existing slicing methods inevitably lead to redundant computation. Therefore, in this article we present a novel adaptive slicing method named ASAHI (Adaptive Slicing Aided Hyper Inference), which can dramatically reduce redundant computation using an adaptive slicing size. Specifically, ASAHI focuses on the number of slices rather than the slicing size, that is, it adaptively adjusts the slicing size to control the number of slices according to the image resolution. Additionally, we replace the standard non-maximum suppression technique with Cluster-DIoU-NMS due to its improved accuracy and inference speed in the post-processing stage. In extensive experiments, ASAHI achieves competitive performance on the VisDrone and xView datasets. The results show that the mAP50 is increased by 0.9% and the computation time is reduced by 20–25% compared with state-of-the-art slicing methods on the TPH-YOLOV5 pretrained model. On the VisDrone2019-DET-val dataset, our mAP50 result is 56.4% higher, demonstrating the superiority of our approach.

In recent years, small object detection has been widely applied to aerial scenes. Although existing small object detection algorithms have achieved significant success, it remains challenging to ensure an acceptable processing speed and detection accuracy simultaneously in high-resolution images. In this work, we propose a novel framework called SliNet, a slicingaided learning network with a SPPFCSPC Block and a ContentAware ReAssembly of Features (CARAFE) Block merged. Since aerial images usually have high resolutions and small detection targets, we slice images into smaller overlapping patches and integrate SPPFCSPC and CARAFE to find attention region with larger receptive field in dense-object scenes. Experimental results on dataset VisDrone-DET2019 show that the SliNet achieves competitive performance and obtains faster detection speed due to the decrease of computational cost. On the VisDrone2019- DET-val dataset, we attain a mAP score of 46.4% with a mAP50 score of 67.1%, which is better than the state-of-the-art models, demonstrating the superiority of our approach

<!-- *** Custom HTML *** --> Cross-modality person re-identification (Re-ID) between RGB and infrared domains is a hot and challenging problem, which aims to retrieve pedestrian images cross-modality and cross-camera views. Since there is a huge gap between two modalities, the difficulty of solving the problem is how to bridge the cross-modality gap with images. However, most approaches solve this issue mainly by increasing interclass discrepancy between features, and few research studies focus on decreasing intraclass cross-modality discrepancy, which is crucial for cross-modality Re-ID. Moreover, we find that despite the huge gap, the attribute representations of the pedestrian are generally unchanged. We provide a different view of the cross-modality person Re-ID problem, which uses additional attribute labels as auxiliary information to increase intraclass cross-modality similarity. First, we manually annotate attribute labels for a large-scale cross-modality Re-ID dataset. Second, we propose an end-to-end network to learn modality-invariant and identity-specific local features with the joint supervision of attribute classification loss and identity classification loss. The experimental results on a large-scale cross-modality Re-ID benchmarks show that our model achieves competitive Re-ID performance compared with the state-of-the-art methods. To demonstrate the versatility of the model, we report the results of our model on the Market-1501 dataset.

Recently, person re-identification has become one of the research hotspots in the field of computer vision and has received extensive attention in the academic community. Inspired by the part-based research of image ReID, this paper presents a novel feature learning and extraction framework for video-based person re-identification, namely, the extended global-local representation learning network (E-GLRN). Given a video sequence of a pedestrian, the holistic and local features are simultaneously extracted using the E-GLRN network. Specifically, for the global feature learning, we adopt the channel attention convolutional neural network (CNN) and the bidirectional long short-term memory (Bi-LSTM) networks, which are responsible for introducing a CNN-LSTM module to learn the features of consecutive frames. The local feature learning module relies on the key local information extraction, which is based on the Bi-LSTM networks. In order to obtain the local feature more effectively, our work defines a concept of “the main image group” by selecting three representative frames. The local feature representation of a video is obtained by exploiting the spatial contextual and appearance information of this group. The local and global features extracted in this paper are complementary and further combined into a discriminative and robust feature representation of the video sequence. Extensive experiments are conducted on three video-based ReID datasets, including the iLIDS-VID, PRID2011 and MARS datasets. The experimental results demonstrate that the proposed method outperforms state-of-the-art video-based re-identification approaches.