Feature-level fusion methods have demonstrated superior performance for visible- infrared object detection due to the deep exploration of visible and infrared features. However, most existing feature-level fusion methods utilize multiple convolutional layers with fixed parameters to extract bimodal features, leading to low adaptivity to diverse data distributions. This paper proposes a Cross-scale Dynamic Convolution-driven YOLO Fusion (CDC-YOLOFusion) network, which introduces a novel Cross-scale Dynamic Convolution Fusion (CDCF) module to adaptively extract and fuse bimodal features concerning on data distribution. Technically, CDC-YOLOFusion first designs a novel data augmentation strategy “Cross-modal Data Swapping” (CDS) to exchange local regions between visible and infrared images, effectively capturing cross-modal correlations within local regions. Building on this, the proposed CDCF utilizes cross-scale enhanced features to assist dynamic convolution prediction by introducing a disparity attention mask, which emphasizes the extraction of disparate features between two modalities. Our CDCF is effectively guided by a novel cross-modal kernel interaction loss, aiming the learned kernels to simultaneously focus on common salient features and unique features of each modality for comprehensive feature generation. Extensive experiments on three representative detection datasets demonstrate that CDCF can be easily plugined into the existing pipelines, obtaining consistent performance improvements. Moreover, our approach yields SOTA performance with about 2% to 3% mAP improvements as compared to the state-of-the-art methods. The code has been published in https://github.com/Morph-3800/CDC_YOLOFusion .

Transformer-based methods have demonstrated superior performance for monocular 3D object detection recently, which aims at predicting 3D attributes from a single 2D image. Most existing transformer-based methods leverage both visual and depth representations to explore valuable query points on objects, and the quality of the learned query points has a great impact on detection accuracy. Unfortunately, existing unsupervised attention mechanisms in transformers are prone to generate low-quality query features due to inaccurate receptive fields, especially on hard objects. To tackle this problem, this paper proposes a novel “Supervised Scale-aware Deformable Attention” (SSDA) for monocular 3D object detection. Specifically, SSDA presets several masks with different scales and utilizes depth and visual features to adaptively learn a scale-aware filter for object query augmentation. Imposing the scale awareness, SSDA could well predict the accurate receptive field of an object query to support robust query feature generation. Aside from this, SSDA is assigned with a Weighted Scale Matching (WSM) loss to supervise scale prediction, which presents more confident results as compared to the unsupervised attention mechanisms. Extensive experiments on the KITTI and Waymo Open datasets demonstrate that SSDA significantly improves the detection accuracy, especially on moderate and hard objects, yielding state-of-the-art performance as compared to the existing approaches. Our code will be made publicly available at https://github.com/mikasa3lili/SSD-MonoDETR .

Visible-infrared object detection has attracted increasing attention recently due to its superior performance and cost-efficiency. Most existing methods focus on the detection of strictly-aligned data, significantly limiting its practical applications. Although several researchers have attempted to explore weakly-aligned visible-infrared object detection, they are limited to small translational deviations and suffer from a low detection speed. This paper first explores non-aligned visibleinfrared object detection with complex deviations in translation, scaling, and rotation, and proposes a fast one-stage detector YOLO-Adaptor, which introduces a lightweight multi-modal adaptor to simultaneously predict alignment parameters and confidence weights between modalities. The adaptor adopts a feature-level alignment during the feature extraction process, ensuring high alignment efficiency. Moreover, we introduce a feature contrastive learning loss to guide the alignment learning of the adaptor, aiming to reduce the representation gap between the two modalities in hyperbolic space to implement feature spatial and distributional consistency. Extensive experiments are conducted on three datasets, including one weakly-aligned and two non-aligned datasets, and the experimental results demonstrate that YOLO-Adaptor could achieve significant performance improvements in terms of speed and accuracy

The cross-domain image captioning, which is trained on a source domain and generalized to other domains, usually faces the large domain shift problem. Although prior work has attempted to leverage both paired source and unpaired target data to minimize this shift, the performance is still unsatisfactory. One main reason lies in the large discrepancy in language expression between two domains, where diverse language styles are adopted to describe an image from different views, resulting in different semantic descriptions for an image. To tackle this problem, this paper proposes a Style-based Cross-domain Image Captioner (SCIC) which incorporates the discriminative style information into the encoder-decoder framework, and interprets an image as a special sentence according to external style instructions. Technically, we design a novel “Instruction-based LSTM”, which adds the instruct gate to collect a style instruction, and then outputs a specified format according to that instruction. Two objectives are designed to train I-LSTM: 1) generating correct image descriptions and 2) generating correct styles, thus the model is expected to accurately capture the semantic meanings of an image by the special caption as well as understand the syntactic structure of the caption. We use MS-COCO as the source domain, and Oxford-102, CUB-200, Flickr30k as the target domains. Experimental results demonstrate that our model consistently outperforms the previous methods, and the style information incorporating with I-LSTM significantly improves the performance, with 5% CIDEr improvements at least on all datasets.

Point-based 3D detection approaches usually suffer from the severe point sampling imbalance problem between foreground and background. We observe that prior works have attempted to alleviate this imbalance by emphasizing foreground sampling. However, even adequate foreground sampling may be extremely unbalanced between nearby and distant objects, yielding unsatisfactory performance in detecting distant objects. To tackle this issue, this paper first proposes a novel method named Distant Object Augmented Set Abstraction and Regression (DO-SA& R) to enhance distant object detection, which is vital for the timely response of decision-making systems like autonomous driving. Technically, our approach first designs DO-SA with novel distant object augmented farthest point sampling (DO-FPS) to emphasize sampling on distant objects by leveraging both object-dependent and depth-dependent information. Then, we propose distant object augmented regression to reweight all the instance boxes for strengthening regression training on distant objects. In practice, the proposed DO-SA&R can be easily embedded into the existing modules, yielding consistent performance improvements, especially on detecting distant objects. Extensive experiments are conducted on the popular KITTI, nuScenes and Waymo datasets, and DO-SA&R demonstrates superior performance, especially for distant object detection. Our code is available at https://github.com/mikasa3lili/DO-SAR.