Streaming Detection of Queried Event Start
Cristóbal Eyzaguirre, Eric Tang, Shyamal Buch, Adrien Gaidon, Jiajun Wu, Juan Carlos Niebles · 2024
Read Streaming Detection of Queried Event Start by Cristóbal Eyzaguirre, Eric Tang, Shyamal Buch, Adrien Gaidon, Jiajun Wu, Juan Carlos Niebles (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-03567.txt.
Description
Cristobal Eyzaguirre and co-authors detect queried event starts in streaming video in a 2024 paper. Robots and smart glasses must react to events as they begin. Users describe the event in plain words. Finding its exact start frame quickly is the challenge.
A new task formalizes that streaming detection problem. A benchmark built from an egocentric video collection supports it. Task-specific metrics reward accuracy and low delay jointly. Adapter-based models transfer image knowledge into video cheaply. Three vision-language backbones were put through the tests. Three adapter designs competed across clip settings. Untrimmed long videos formed the harder evaluation. Short clips formed the controlled comparison. Parameter-efficient tuning inspired the adapter approach. Online modeling must stay light for real deployment. The benchmark invites future systems to beat these baselines. Embodied applications are the end users envisioned.
Catalog record
| CREATOR | Cristóbal Eyzaguirre, Eric Tang, Shyamal Buch, Adrien Gaidon, Jiajun Wu, Juan Carlos Niebles |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 106 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-03567