Complex video queries can be answered by decomposing them into modular subtasks. However, existing video data management systems assume the existence of predefined modules for each subtask. We introduce VOCAL-UDF, a novel self-enhancing system that supports compositional queries over videos without the need for predefined modules. VOCAL-UDF automatically identifies and constructs missing modules and encapsulates them as user-defined functions (UDFs), thus expanding its querying capabilities. To achieve this, we formulate a unified UDF model that leverages large language models (LLMs) to aid in new UDF generation. VOCAL UDF handles a wide range of concepts by supporting both program-based UDFs (i.e., Python functions generated by LLMs) and distilled-model UDFs (lightweight vision models distilled from strong pretrained models). To resolve the inherent ambiguity in user intent, VOCAL-UDF generates multiple candidate UDFs and uses active learning to efficiently select the best one. With the self-enhancing capability, VOCAL-UDF significantly improves query performance across three video datasets.

Large language models (LLMs) are powerful tools for content moderation, but their inference costs and latency make them prohibitive for casual use on large datasets, such as the Google Ads repository. This study proposes a method for scaling up LLM reviews for content moderation in Google Ads. First, we use heuristics to select candidates via filtering and duplicate removal, and create clusters of ads for which we select one representative ad per cluster. We then use LLMs to review only the representative ads. Finally, we propagate the LLM decisions for the representative ads back to their clusters. This method reduces the number of reviews by more than 3 orders of magnitude while achieving a 2x recall compared to a baseline non-LLM model. The success of this approach is a strong function of the representations used in clustering and label propagation; we found that cross-modal similarity representations yield better results than uni-modal representations.

We introduce VOCALExplore, a system designed to support users in building domain-specific models over video datasets. VOCALExplore supports interactive labeling sessions and trains models using user-supplied labels. VOCALExplore maximizes model quality by automatically deciding how to select samples based on observed skew in the collected labels. It also selects the optimal video representations to use when training models by casting feature selection as a rising bandit problem. Finally, VOCALExplore implements optimizations to achieve low latency without sacrificing model performance. We demonstrate that VOCALExplore achieves close to the best possible model quality given candidate acquisition functions and feature extractors, and it does so with low visible latency (~1 second per iteration) and no expensive preprocessing.

We demonstrate EQUI-VOCAL, a system that synthesizes compositional queries over videos from user feedback. EQUI-VOCAL enables users to query a video database for complex events by providing a few positive and negative examples of what they are looking for and labeling a small number of additional system-selected examples. Using those user inputs, EQUI-VOCAL synthesizes declarative queries that can then retrieve additional instances of the desired events. The demonstration makes two contributions: it introduces EQUI-VOCAL's graphical user interface and enables conference attendees to experiment with EQUI-VOCAL on a variety of queries. Both enable users to gain a better understanding of EQUI-VOCAL's query synthesis approach and to explore the impact of hyperparameters and label noise on system performance.

We introduce EQUI-VOCAL: a new system that automatically synthesizes queries over videos from limited user interactions. The user only provides a handful of positive and negative examples of what they are looking for. EQUI-VOCAL utilizes these initial examples and additional ones collected through active learning to efficiently synthesize complex user queries. Our approach enables users to find events without database expertise, with limited labeling effort, and without declarative specifications or sketches. Core to EQUI-VOCAL's design is the use of spatio-temporal scene graphs in its data model and query language and a novel query synthesis approach that works on large and noisy video data. Our system outperforms two baseline systems---in terms of F1 score, synthesis time, and robustness to noise---and can flexibly synthesize complex queries that the baselines do not support.

To support the massive data requirements of modern supervised machine learning (ML) algorithms, crowdsourcing systems match volunteer contributors to appropriate tasks. Such systems learn what types of tasks contributors are interested to complete. In this paper, instead of focusing on what to ask, we focus on learning how to ask: how to make relevant and interesting requests to encourage crowdsourcing participation. We introduce a new technique that augments questions with ML-based request strategies drawn from social psychology. We also introduce a contextual bandit algorithm to select which strategy to apply for a given task and contributor. We deploy our approach to collect volunteer data from Instagram for the task of visual question answering (VQA), an important task in computer vision and natural language processing that has enabled numerous human-computer interaction applications. For example, when encountering a user’s Instagram post that contains the ornate Trevi Fountain in Rome, our approach learns to augment its original raw question “Where is this place?” with image-relevant compliments such as “What a great statue!” or with travel-relevant justifications such as “I would like to visit this place”, increasing the user’s likelihood of answering the question and thus providing a label. We deploy our agent on Instagram to ask questions about social media images, finding that the response rate improves from 15.8% with unaugmented questions to 30.54% with baseline rule-based strategies and to 58.1% with ML-based strategies.