Class-incremental learning is one of the most important settings for the study of Continual Learning, as it closely resembles real-world application scenarios. With constrained memory sizes, catastrophic forgetting ar...
详细信息
In the classical supervised learning settings, classifiers are fit with the assumption of balanced label distributions and produce remarkable results on the same. In the real world, however, these assumptions often be...
详细信息
As the use of deep neural networks continues to grow, understanding their behaviour has become more crucial than ever. Post-hoc explainability methods are a potential solution, but their reliability is being called in...
详细信息
Rip currents are the leading cause of fatal accidents and injuries on many beaches worldwide, emphasizing the importance of automatically detecting these hazardous surface water currents. In this paper, we address a n...
详细信息
Although fast adversarial training provides an efficient approach for building robust networks, it may suffer from a serious problem known as catastrophic overfitting (CO), where multi-step robust accuracy suddenly co...
详细信息
Generating grammatically and semantically correct captions in video captioning is a challenging task.. The captions generated from the existing methods are either word-by-word that do not align with grammatical struct...
详细信息
In the past few years, Siamese networks have achieved outstanding improvements in visual object tracking. However, visual distractors with similar semantics can be easily misclassified as the target by Siamese network...
详细信息
ISBN:
(数字)9781665487399
ISBN:
(纸本)9781665487399
In the past few years, Siamese networks have achieved outstanding improvements in visual object tracking. However, visual distractors with similar semantics can be easily misclassified as the target by Siamese networks and may consequently result in the drift problem. Besides, the Hanning window penalty, which is generally used to suppress distractors, could fail in many challengeable scenes. Notably, most failures violate the assumption of motion continuity. Thus, in this work, we explore motion information to mitigate the drift problem in visual tracking. First, we introduce a simple linear Kalman filter to predict the bounding box of the target in the current frame, which acts as a reference for decisions. Second, an IoU-Guided penalty is assembled in the post-processing to suppress distractors effectively. It's worth mentioning that our method is almost cost-free. We conduct numerous experimental validations and analyses of our approach on several challenging sequences and datasets. Our tracker runs at approximately 40 fps and performs well on those sequences which include the Background Clutter attribute. Finally, by simultaneously integrating the IoU-Guidedpenalty and the Hanning window penalty with a strong baseline tracker TransT [7], our method achieves favorable gains by 69.1 -> 71.5, 65.7 -> 66.7, 64.9 -> 65.9 success on OTB-100 [32], LaSOT [12], NFS[18].
With the increasing popularity and the increasing size of vision transformers (ViTs), there has been an increasing interest in making them more efficient and less computationally costly for deployment on edge devices ...
详细信息
Anticipation problem has been studied considering different aspects such as predicting humans' locations, predicting hands and objects trajectories, and forecasting actions and human-object interactions. In this p...
详细信息
The retail sector has experienced significant growth in artificial intelligence and computervision applications, particularly with the emergence of automatic checkout (ACO) systems in stores and supermarkets. ACO sys...
详细信息
暂无评论