Strengthening the inspection of outfalls into rivers and oceans can help monitor pollutant emissions to the natural environment. Unmanned aerial vehicle (UAV) with high spatial resolution imagery has become a more efficient method for outfall surveys. At present, outfalls retrieval from UAV images relies on visual interpretation by skilled experts. However, long periods of concentration on detecting outfalls in high-resolution images for an expert easily increase mental load and stress, resulting in missing and false detection. Therefore, we develop a deep learning model, called Attention Based Small Object Detector (ABSOD), to perform outfalls detection in aerial images. In this model, an adaptive spatial correlation pyramid attention (ASCPA) network is proposed to establish long-distance region-to-region relationships between the outfall and its surrounding information more effectively. This network is mainly composed of SPE (Spatial Pyramid Extractor) and SCFM (Spatial Correlation Fusion Module). The purpose of the SPE is to extract multi-scale spatial information on the feature map. The SCFM is used to perform spatial correlation feature recalibration to selectively emphasized informative features. Experimental results show that the proposed network outperforms the state-of-the-art small object detection model in detecting outfalls, and reaches 45.9%, 92.8%, 86.5% and 34.4% in the four metrics of Precision, Recall, AP 0.5 , and AP 0.5:0.95 , respectively. To show the superiority of the ASCPA network, we compared our results with other attention mechanisms, all of them show that the ASCPA network has a competitive performance for outfalls detection. Moreover, based on visualization analysis, the ASCPA network is able to pay more attention on true outfall objects with respect to other attention mechanisms. These promising results demonstrate that the deep learning algorithm can be a feasible solution to assist experts in detecting outfalls with UAV imagery. The model and code are available at https://github.com/ISCLab-Bistu/ASCPA-Attention .

In the field of autonomous driving, monocular 3D detection is a critical task which estimates 3D properties (depth, dimension, and orientation) of objects in a single RGB image. Previous works have used features in a heuristic way to learn 3D properties, without considering that inappropriate features could have adverse effects. In this paper, sample selection is introduced that only suitable samples should be trained to regress the 3D properties. To select samples adaptively, we propose a Learnable Sample Selection (LSS) module, which is based on Gumbel-Softmax and a relative-distance sample divider. The LSS module works under a warm-up strategy leading to an improvement in training stability. Additionally, since the LSS module dedicated to 3D property sample selection relies on object-level features, we further develop a data augmentation method named MixUp3D to enrich 3D property samples which conforms to imaging principles without introducing ambiguity. As two orthogonal methods, the LSS module and MixUp3D can be utilized independently or in conjunction. Sufficient experiments have shown that their combined use can lead to synergistic effects, yielding improvements that transcend the mere sum of their individual applications. Leveraging the LSS module and the MixUp3D, without any extra data, our method named MonoLSS ranks 1st in all three categories (Car, Cyclist, and Pedestrian) on KITTI 3D object detection benchmark, and achieves competitive results on both the Waymo dataset and KITTI-nuScenes cross-dataset evaluation. The code is included in the supplementary material and will be released to facilitate related academic and industrial studies.

Monocular 3D detection of vehicle and infrastructure sides are two important topics in autonomous driving. Due to diverse sensor installations and focal lengths, researchers are faced with the challenge of constructing algorithms for the two topics based on different prior knowledge. In this paper, by taking into account the diversity of pitch angles and focal lengths, we propose a unified optimization target named normalized depth, which realizes the unification of 3D detection problems for the two sides. Furthermore, to enhance the accuracy of monocular 3D detection, 3D normalized cube depth of obstacle is developed to promote the learning of depth information. We posit that the richness of depth clues is a pivotal factor impacting the detection performance on both the vehicle and infrastructure sides. A richer set of depth clues facilitates the model to learn better spatial knowledge, and the 3D normalized cube depth offers sufficient depth clues. Extensive experiments demonstrate the effectiveness of our approach. Without introducing any extra information, our method, named MonoUNI, achieves state-of-the-art performance on five widely used monocular 3D detection benchmarks, including Rope3D and DAIR-V2X-I for the infrastructure side, KITTI and Waymo for the vehicle side, and nuScenes for the cross-dataset evaluation.