Concerns for the privacy of individuals captured in public imagery have led to privacy-preserving action recognition. Existing approaches often suffer from issues arising through obfuscation being applied globally and a lack of interpretability. Global obfuscation hides privacy sensitive regions, but also contextual regions important for action recognition. Lack of interpretability erodes trust in these new technologies. We highlight the limitations of current paradigms and propose a solution: Human selected privacy templates that yield interpretability by design, an ob-fuscation scheme that selectively hides attributes and also induces temporal consistency, which is important in action recognition. Our approach is architecture agnostic and directly modifies input imagery, while existing approaches generally require architecture training. Our approach offers more flexibility, as no training is required, and outperforms alternatives on three widely used datasets.

Multiscale video transformers have been explored in a wide variety of vision tasks. To date, however, the multiscale processing has been confined to the encoder or decoder alone. We present a unified multiscale encoder-decoder transformer that is focused on dense prediction tasks in videos. Multiscale representation at both encoder and decoder yields key benefits of implicit extraction of spatiotemporal features (i.e. without reliance on input optical flow) as well as temporal consistency at encoding and coarse-to-fine detection for high-level (e.g. object) semantics to guide precise localization at decoding. Moreover, we propose a transductive learning scheme through many-to-many label propagation to provide temporally consistent predictions. We showcase our Multiscale Encoder-Decoder Video Transformer (MED-VT) on Automatic Video Object Segmentation (AVOS) and actor/action segmentation, where we outperform state-of-the-art approaches on multiple benchmarks using only raw images, without using optical flow.

In this paper, we study the problem of procedure planning in instructional videos. Here, an agent must produce a plausible sequence of actions that can transform the environment from a given start to a desired goal state. When learning procedure planning from instructional videos, most recent work leverages intermediate visual observations as supervision, which requires expensive annotation efforts to localize precisely all the instructional steps in training videos. In contrast, we remove the need for expensive temporal video annotations and propose a weakly supervised approach by learning from natural language instructions. Our model is based on a transformer equipped with a memory module, which maps the start and goal observations to a sequence of plausible actions. Furthermore, we augment our model with a probabilistic generative module to capture the uncertainty inherent to procedure planning, an aspect largely overlooked by previous work. We evaluate our model on three datasets and show our weakly-supervised approach outperforms previous fully supervised state-of-the-art models on multiple metrics.

Goal-conditioned approaches recently have been found very useful to human trajectory prediction, when adequate goal estimates are provided. Yet, goal inference is difficult in itself and often incurs extra learning effort. We propose to predict pedestrian trajectories via the guidance of goal expertise, which can be obtained with modest expense through a novel goal-search mechanism on already seen training examples. There are three key contributions in our study. First, we devise a framework that exploits nearest examples for high-quality goal position inquiry. This approach naturally considers multi-modality, physical constraints, compatibility with existing methods and is nonparametric; it therefore does not require additional learning effort typical in goal inference. Second, we present an end-to-end trajectory predictor that can efficiently associate goal retrievals to past motion information and dynamically infer possible future trajectories. Third, with these two novel techniques in hand, we conduct a series of experiments on two broadly explored datasets (SDD and ETH/UCY) and show that our approach surpasses previous state-of-the-art performance by notable margins and reduces the need for additional parameters. Code can be found at our project page1.

Recognizing actions from limited preliminary video observations has seen considerable recent progress. Typically, however, such progress has been had without explicitly modeling fine-grained motion evolution as a potentially valuable information source. In this study, we address this task by investigating how action patterns evolve over time in a spatial feature space. There are three key components to our system. First, we work with intermediate-layer ConvNet features, which allow for abstraction from raw data, while retaining spatial layout, which is sacrificed in approaches that rely on vectorized global representations. Second, instead of propagating features per se, we propagate their residuals across time, which allows for a compact representation that reduces redundancy while retaining essential information about evolution over time. Third, we employ a Kalman filter to combat error build-up and unify across prediction start times. Extensive experimental results on the JHMDB21, UCF101 and BIT datasets show that our approach leads to a new state-of-the-art in action prediction.