Object detection in drone-captured images is a popular task in recent years. As drones always navigate at different altitudes, the object scale varies considerably, which burdens the optimization of models. Moreover, high-speed and low-altitude flight cause motion blur on densely packed objects, which leads to great challenges. To solve the two issues mentioned above, based on YOLOv5, we add an additional prediction head to detect tiny-scale objects and replace CNN-based prediction heads with transformer prediction heads (TPH), constructing the TPH-YOLOv5 model. TPH-YOLOv5++ is proposed to significantly reduce the computational cost and improve the detection speed of TPH-YOLOv5. In TPH-YOLOv5++, cross-layer asymmetric transformer (CA-Trans) is designed to replace the additional prediction head while maintain the knowledge of this head. By using a sparse local attention (SLA) module, the asymmetric information between the additional head and other heads can be captured efficiently, enriching the features of other heads. In the VisDrone Challenge 2021, TPH-YOLOv5 won 4th place and achieved well-matched results with the 1st place model (AP 39.43%). Based on the TPH-YOLOv5 and CA-Trans module, TPH-YOLOv5++ can further increase efficiency while achieving comparable and better results.

Few-shot segmentation focuses on the generalization of models to segment unseen object with limited annotated samples. However, existing approaches still face two main challenges. First, huge feature distinction between support and query images causes knowledge transferring barrier, which harms the segmentation performance. Second, limited support prototypes cannot adequately represent features of support objects, hard to guide high-quality query segmentation. To deal with the above two issues, we propose self-distillation embedded supervised affinity attention model to improve the performance of few-shot segmentation task. Specifically, the self-distillation guided prototype module uses self-distillation to align the features of support and query. The supervised affinity attention module generates high-quality query attention map to provide sufficient object information. Extensive experiments prove that our model significantly improves the performance compared to existing methods. Comprehensive ablation experiments and visualization studies also show the significant effect of our method on few-shot segmentation task. On COCO-20i dataset, we achieve new state-of-the-art results. Training code and pretrained models are available at https://github.com/cv516Buaa/SD-AANet.

Visual navigation, characterized by its autonomous capabilities, cost effectiveness, and robust resistance to interference, serves as the foundation for vision-based autonomous landing systems. These systems rely heavily on runway instance segmentation, which accurately divides runway areas and provides precise information for unmanned aerial vehicle (UAV) navigation. However, current research primarily focuses on runway detection but lacks relevant runway instance segmentation datasets. To address this research gap, we created the Runway Landing Dataset (RLD), a benchmark dataset that focuses on runway instance segmentation mainly based on X-Plane. To overcome the challenges of large-scale changes and input image angle differences in runway instance segmentation tasks, we propose a vision-based autonomous landing segmentation network (VALNet) that uses band-pass filters, where a Context Enhancement Module (CEM) guides the model to learn adaptive “band” information through heatmaps, while an Orientation Adaptation Module (OAM) of a triple-channel architecture to fully utilize rotation information enhances the model’s ability to capture input image rotation transformations. Extensive experiments on RLD demonstrate that the new method has significantly improved performance. The visualization results further confirm the effectiveness and interpretability of VALNet in the face of large-scale changes and angle differences. This research not only advances the development of runway instance segmentation but also highlights the potential application value of VALNet in vision-based autonomous landing systems. Additionally, RLD is publicly available.

Recent CNNs (convolutional neural networks) have become more and more compact. The elegant structure design highly improves the performance of CNNs. With the development of knowledge distillation technique, the performance of CNNs gets further improved. However, existing knowledge distillation guided methods either rely on offline pretrained high-quality large teacher models or online heavy training burden. To solve the above problems, we propose a feature-sharing and weight-sharing based ensemble network (training framework) guided by knowledge distillation (EKD-FWSNet) to make baseline models stronger in terms of representation ability with less training computation and memory cost involved. Specifically, motivated by getting rid of the dependence of offline pretrained teacher model, we design an end-to-end online training scheme to optimize EKD-FWSNet. Motivated by decreasing the online training burden, we only introduce one auxiliary classmate branch to construct multiple forward branches, which will then be integrated as ensemble teacher to guide baseline model. Compared to previous online ensemble training frameworks, EKD-FWSNet can provide diverse output predictions without relying on increasing auxiliary classmate branches. Motivated by maximizing the optimization power of EKD-FWSNet, we exploit the representation potential of weight-sharing blocks and design efficient knowledge distillation mechanism in EKD-FWSNet. Extensive comparison experiments and visualization analysis on benchmark datasets (CIFAR-10/100, tiny-ImageNet, CUB-200 and ImageNet) show that self-learned EKD-FWSNet can boost the performance of baseline models by large margin, which has obvious superiority compared to previous related methods. Extensive analysis also proves the interpretability of EKD-FWSNet. Our code is available at https://github.com/cv516Buaa/EKD-FWSNet .

Airport detection in remote sensing scenes is a crucial area of research, playing a key role in aircraft blind landing procedures. However, airport detection in remote sensing scenes still faces challenges such as class confusion, poor detection performance on multi-scale objects, and limited dataset availability. To address these issues, this paper proposes a novel airport detection network (TPH-YOLOv5-Air) based on adaptive spatial feature fusion (ASFF). Firstly, we construct an Airport Confusing Object Dataset (ACD) specifically tailored for remote sensing scenarios containing 9501 instances of airport confusion objects. Secondly, building upon the foundation of TPH-YOLOv5++, we adopt the ASFF structure, which not only enhances the feature extraction efficiency but also enriches feature representation. Moreover, an adaptive spatial feature fusion (ASFF) strategy based on adaptive parameter adjustment module (APAM) is proposed, which improves the feature scale invariance and enhances the detection of airports. Finally, experimental results based on the ACD dataset demonstrate that TPH-YOLOv5-Air achieves a mean average precision (mAP) of 49.4%, outperforming TPH-YOLOv5++ by 2% and the original YOLOv5 network by 3.6%. This study contributes to the advancement of airport detection in remote sensing scenes and demonstrates the practical application potential of TPH-YOLOv5-Air in this domain. Visualization and analysis further validate the effectiveness and interpretability of TPH-YOLOv5-Air. The ACD dataset is publicly available.