Out-of-distribution (OOD) detection remains a fundamental challenge for deep neural networks, particularly due to overconfident predictions on unseen OOD samples during testing. We reveal a key insight: OOD samples predicted as the same class, or given high probabilities for it, are visually more similar to each other than to the true in-distribution (ID) samples. Motivated by this class-specific observation, we propose DCAC (Dynamic Class-Aware Cache), a training-free, test-time calibration module that maintains separate caches for each ID class to collect high-entropy samples and calibrate the raw predictions of input samples. DCAC leverages cached visual features and predicted probabilities through a lightweight two-layer module to mitigate overconfident predictions on OOD samples. This module can be seamlessly integrated with various existing OOD detection methods across both unimodal and vision-language models while introducing minimal computational overhead. Extensive experiments on multiple OOD benchmarks demonstrate that DCAC significantly enhances existing methods, achieving substantial improvements, i.e., reducing FPR95 by 6.55% when integrated with ASH-S on ImageNet OOD benchmark.

To reduce the efforts of human in browsing long surveillance videos, synopsis videos are proposed. Traditional synopsis video generation methods condense most of the activities in the video by simultaneously showing several actions, even when they originally occurred at different times. This inevitably causes ignorance of temporal relationship among objects. For example, two persons walk shoulder to shoulder and they are detected and tracked separately, but in the synopsis they never ‘met’. In this study, a trajectory mapping model is defined, whose energy function includes not only the cost caused by the synopsis video, but that of the original video. In this way, it tries to make the relationship between objects of the original video consistent with that of the synopsis. Finally, the video synopsis is generated by an energy minimisation method. Experiments show that the proposed video synopsis can reduce the spatiotemporal redundancies of the input video as much as possible. Moreover, it can keep the important relationship between objects and maintain the time consistency of important activities.