An official website of the United States government
Multi-camera systems are essential in movies, live broadcasts, and other media. The selection of the appropriate camera for every moment has a decisive impact on production quality and audience preferences. Learning-based multi-camera view recommendation frameworks have been explored to assist professionals in decision making. This work explores how two standard cinematography practices could be incorporated into the learning pipeline: (1) not staying on the same camera for too long and (2) introducing a scene from a wider shot and gradually progressing to narrower ones. In these regards, we incorporate (1) the duration of the displaying camera and (2) camera identity as temporal and camera embedding in a transformer architecture, thereby implicitly guiding the model to learn the two practices from professional-labeled data. Experiments show that the proposed framework outperforms the baseline by 14.68% in six-way classification accuracy. Ablation studies on different approaches to embedding the temporal and camera information further verify the efficacy of the framework.
more »
« less
This content will become publicly available on December 8, 2025
Pseudo Dataset Generation for Out-of-domain Multi-Camera View Recommendation
Multi-camera systems are indispensable in movies, TV shows, and other media. Selecting the appropriate camera at every timestamp has a decisive impact on production quality and audience preferences. Learning-based view recommendation frameworks can assist professionals in decision-making. However, they often struggle outside of their training domains. The scarcity of labeled multi-camera view recommendation datasets exacerbates the issue. Based on the insight that many videos are edited from the original multi-camera videos, we propose transforming regular videos into pseudo-labeled multi-camera view recommendation datasets. Promisingly, by training the model on pseudo-labeled datasets stemming from videos in the target domain, we achieve a 68% relative improvement in the model’s accuracy in the target domain and bridge the accuracy gap between in-domain and never-before-seen domains.
more »
« less
- PAR ID:
- 10635556
- Publisher / Repository:
- IEEE
- Date Published:
- ISBN:
- 979-8-3315-2954-3
- Page Range / eLocation ID:
- 1 to 5
- Subject(s) / Keyword(s):
- Training Accuracy TV Visual communication Decision making Production Cinematography Media Motion pictures Videos
- Format(s):
- Medium: X
- Location:
- Tokyo, Japan
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
The scarcity of labeled data has traditionally been the primary hindrance in building scalable supervised deep learning models that can retain adequate performance in the presence of various heterogeneities in sample distributions. Domain adaptation tries to address this issue by adapting features learned from a smaller set of labeled samples to that of the incoming unlabeled samples. The traditional domain adaptation approaches normally consider only a single source of labeled samples, but in real world use cases, labeled samples can originate from multiple-sources – providing motivation for multi-source domain adaptation (MSDA). Several MSDA approaches have been investigated for wearable sensor-based human activity recognition (HAR) in recent times, but their performance improvement compared to single source counterpart remained marginal. To remedy this performance gap that, we explore multiple avenues to align the conditional distributions in addition to the usual alignment of marginal ones. In our investigation, we extend an existing multi-source domain adaptation approach under semi-supervised settings. We assume the availability of partially labeled target domain data and further explore the pseudo labeling usage with a goal to achieve a performance similar to the former. In our experiments on three publicly available datasets, we find that a limited labeled target domain data and pseudo label data boost the performance over the unsupervised approach by 10-35% and 2-6%, respectively, in various domain adaptation scenarios.more » « less
-
Given a multi-view video, which viewpoint is most informative for a human observer? Existing methods rely on heuristics or expensive "best-view" supervision to answer this question, limiting their applicability. We propose a weakly supervised approach that leverages language accompanying an instructional multi-view video as a means to recover its most informative viewpoint(s). Our key hypothesis is that the more accurately an individual view can predict a view-agnostic text summary, the more informative it is. To put this into action, we propose LangView, a framework that uses the relative accuracy of view-dependent caption predictions as a proxy for best view pseudo-labels. Then, those pseudo-labels are used to train a view selector, together with an auxiliary camera pose predictor that enhances view-sensitivity. During inference, our model takes as input only a multi-view video--no language or camera poses--and returns the best viewpoint to watch at each timestep. On two challenging datasets comprised of diverse multi-camera setups and how-to activities, our model consistently outperforms state-of-the-art baselines, both with quantitative metrics and human evaluation.more » « less
-
We propose XVO, a semi-supervised learning method for training generalized monocular Visual Odometry (VO) models with robust off-the-self operation across diverse datasets and settings. In contrast to standard monocular VO approaches which often study a known calibration within a single dataset, XVO efficiently learns to recover relative pose with real-world scale from visual scene semantics, i.e., without relying on any known camera parameters. We optimize the motion estimation model via self-training from large amounts of unconstrained and heterogeneous dash camera videos available on YouTube. Our key contribution is twofold. First, we empirically demonstrate the benefits of semi-supervised training for learning a general-purpose direct VO regression network. Second, we demonstrate multi-modal supervision, including segmentation, flow, depth, and audio auxiliary prediction tasks, to facilitate generalized representations for the VO task. Specifically, we find audio prediction task to significantly enhance the semi-supervised learning process while alleviating noisy pseudo-labels, particularly in highly dynamic and out-of-domain video data. Our proposed teacher network achieves state-of-the-art performance on the commonly used KITTI benchmark despite no multi-frame optimization or knowledge of camera parameters. Combined with the proposed semi-supervised step, XVO demonstrates off-the-shelf knowledge transfer across diverse conditions on KITTI, nuScenes, and Argoverse without fine-tuning.more » « less
-
We address the problem of human action classification in drone videos. Due to the high cost of capturing and labeling large-scale drone videos with diverse actions, we present unsupervised and semi-supervised domain adaptation approaches that leverage both the existing fully annotated action recognition datasets and unannotated (or only a few annotated) videos from drones. To study the emerging problem of drone-based action recognition, we create a new dataset, NEC-DRONE, containing 5,250 videos to evaluate the task. We tackle both problem settings with 1) same and 2) different action label sets for the source (e.g., Kinectics dataset) and target domains (drone videos). We present a combination of video and instance-based adaptation methods, paired with either a classifier or an embedding-based framework to transfer the knowledge from source to target. Our results show that the proposed adaptation approach substantially improves the performance on these challenging and practical tasks. We further demonstrate the applicability of our method for learning cross-view action recognition on the Charades-Ego dataset. We provide qualitative analysis to understand the behaviors of our approaches.more » « less
