Vehicle counting from an unmanned aerial vehicle (UAV) is becoming a popular research topic in traffic monitoring. Camera mounted on UAV can be regarded as a visual sensor for collecting aerial videos. Compared with traditional sensors, the UAV can be flexibly deployed to the areas that need to be monitored and can provide a larger perspective. In this paper, a novel framework for vehicle counting based on aerial videos is proposed. In our framework, the moving-object detector can handle the following two situations: static background and moving background. For static background, a pixel-level video foreground detector is given to detect vehicles, which can update background model continuously. For moving background, image-registration is employed to estimate the camera motion, which allows the vehicles to be detected in a reference coordinate system. In addition, to overcome the change of scale and shape of vehicle in images, we employ an online-learning tracker which can update the samples used for training. Finally, we design a multi-object management module which can efficiently analyze and validate the status of the tracked vehicles with multi-threading technique. Our method was tested on aerial videos of real highway scenes that contain fixed-background and moving-background. The experimental results show that the proposed method can achieve more than 90% and 85% accuracy of vehicle counting in fixed-background videos and moving-background videos respectively.

Person re-identification is an important surveillance task of searching and identifying pedestrian across different images or video frames. Despite a significant progress has been made in person re-identification based on RGB image sensors, few work focus on the person re-identification between RGB and infrared images, which is a challenging cross-modality problem and has been widely encountered in a dark environment or at night. In addition to the challenges for the same identity associated with variations in camera viewpoints and person poses, there is a non-negligible shift across different sensor modalities since the visual characteristics from RGB and infrared images are heterogeneous. In this paper, we propose a novel end-to-end dual-path multi-branch network for RGB-infrared cross-modality person re-identification, which introduces the multi-branch deep network architecture. The experimental results obtained with SYSU-MM01 datasets indicate that the proposed method can successfully transfer descriptive visual characteristic between RGB and infrared sensor modality. It can significantly outperform state-of-the-art conventional methods and convolutional neural network methods.

Recently, optical flow estimation benefits greatly from deep learning based techniques. Most approaches use encoder-decoder architecture (U-Net) or spatial pyramid network (SPN) to learn optical flow. Both U-Net and SPN can extract multi-scale features and can predict optical flow directly. However, existing networks ignore to exploit the global information among channel features and inter-spatial relationship of features. In this paper, we propose a dual self-attention pyramid network, which adaptively integrates local features with their global dependencies and focuses on important features and suppresses unimportant features. Specifically, we introduce two types of attention modules into SPN, which emphasizes meaningful features along channel and spatial axes. The channel attention can adaptively re-weight channel-wise features by considering interdependencies among channels. Moreover, the spatial attention can utilize global contextual information to emphasize or suppress features in different spatial locations. In addition, two attention modules are embedded into each pyramidal level, which can refine features at different scale. We evaluate our method on MPI-Sintel and KITTI. The experimental results show that using the dual self-attention module can improve the representation power of network and further increase the accuracy of optical flow estimation.

Supervised depth prediction and optical flow estimation have achieved promising performance due to the advanced deep network architectures. Since the ground truths are difficult to be collected, many recent works try to learn the depth and flow in an unsupervised manner. However, existing methods only use features from convolutional layers or a simple aggregation of multi-level features to predict the depth and flow maps, which is insufficient to exploit context information. In this paper, we attempt to exploit object contextual information and investigate the effect of the object context for joint learning of depth and optical flow. Specifically, we present a novel combination of object context and the framework of joint learning depth and optical flow. Our proposed network can exploit and integrate the object context for both tasks by aggregating the context according to pair-wise similarities. Furthermore, we adopt the existing spatial pyramid network (SPN) to estimate the depth and flow in a coarse-to-fine strategy effectively. Given temporally adjacent stereo pairs, our network can be trained end-to-end in an unsupervised manner and can predict the depth and flow maps simultaneously. We conduct experiments on two publicly available datasets, KITTI2012 and KITTI2015. Our proposed approach yields comparable performance on both depth and flow tasks, compared to the recent deep learning-based approaches. Experimental results demonstrate that exploiting object contextual information is useful and beneficial for depth and optical flow estimation.