Suivi multiobjet à l'aide de l'apprentissage automatique et de points au haut des objets
Bibliographic record
Abstract
RÉSUMÉ: Le suivi multiobjet est un domaine de recherche scientifique attrayant en raison des ses nombreuses applications académiques et commerciales. Cette tâche consiste à détecter les objets de classes prédéterminées à chaque trame d’une séquence vidéo, puis de faire l’association temporelle des détections afin de créer un tracklet pour chaque objet afin de modéliser son mouvement à travers la séquence vidéo. Le but est donc de concevoir une méthode qui parvient à créer les tracklets les plus fidèles à la réalité que possible. Les récentes avancées en apprentissage automatique ont permis de grandement améliorer les méthodes à l’état de l’art, mais il reste encore place à l’amélioration. En effet, un problème majeur qui doit toujours être surmonté est la gestion des occlusions. Les occlusions peuvent être catégorisées comme étant partielles ou complètes. D’abord, lorsqu’un objet est partiellement caché, il devient plus difficile à détecter à cause de la perte d’information que l’occlusion engendre. Il faut donc que notre méthode soit capable d’extraire des caractéristiques assez robustes afin d’être en mesure de détecter les objets même lorsqu’ils sont partiellement cachés. Puis, lorsqu’un objet est complètement caché, il devient impossible de le détecter avec les outils conventionnels, alors il est nécessaire d’adopter des stratégies spécifiquement conçues pour prédire son mouvement ou pour récupérer son tracklet une fois que l’objet redevient visible si l’on veut réussir à continuer de le suivre. ABSTRACT: Multi-object tracking has been a popular field of research in recent years because of its many academic and commercial uses. It consists of detecting objects of pre-defined classes at every frame of a video sequence and then associating them through time in order to create tracklets that model the trajectories of each object. The more the tracklets correspond to the actual trajectory of the objects, the better. Although, the field of multi-object tracking has existed for decades, recent advancements made in deep learning significantly improved the results achieved by tracking models. However, there are still many problems that need to be solved. One of the biggest problems that still plagues multi-object tracking is how to deal with partial and full occlusions. When an object is partially occluded, it becomes a lot harder to detect because of the loss of information incurred by the part of the object that is hidden. In these cases, it becomes extremely important that our method extracts robust and reliable features in order to be still able to detect the objects. When an object is fully occluded, it becomes impossible to detect it through conventional means, so it becomes necessary to use specially-crafted tools that can predict the movement of the object in order to follow it while it is fully occluded or that can link it back to its previous identity once it reappears.
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 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.003 | 0.003 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.009 | 0.004 |
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".