AVQS: Advanced Video Querying System for Machine Learning Applications
Bibliographic record
Abstract
The increasing volume of video data, exemplified by datasets such as YouTube-8M, presents significant challenges to data engineering processes, particularly in efficiently extracting relevant information for model training. Existing video querying systems often fall short in providing a comprehensive solution, lacking the ability to fetch relevant video sections and frames, analyze videos for effective searching, and store data in a centralized location. These limitations are particularly problematic for enterprises with large datasets, often composed of multiple sources, where new projects frequently emerge. Such projects require specific types of videos, such as those featuring cars or animals. Currently, meeting these changing needs requires reprocessing existing datasets or relying on previously assigned tags, which may not capture the full range of complex interactions. In response to these challenges, we propose a video querying system that enables precise searches within large-scale video datasets. Our solution employs Yolo11 for object detection and tracking, generating detailed metadata that includes bounding box coordinates and timestamps. This metadata is stored in a MongoDB database, allowing users to perform complex queries, such as identifying when a person and a car are within 10 pixels of each other. By optimizing the retrieval of relevant video segments and enhancing the tagging process, our system aims to meet the evolving needs of enterprises while leveraging the current technological capabilities of the Yolo11 model.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.004 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.003 | 0.002 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.019 | 0.012 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".