Just a sanity check, for the NLQ task, do we process the entire clip? I see that there is a clip_start_sec and clip_end_sec associated with every clip. Does this mean we might sometimes only use part of the clip for the NLQ queries? Thanks!
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| NLQ Annotation Zero Temporal Windows | 5 | 649 | August 15, 2022 | |
| Ambiguity in NLQ annotations | 2 | 337 | May 5, 2023 | |
| How do we deal with multiple annotations for the Moments Queries task? | 2 | 442 | July 20, 2022 | |
| Ego4d clip names not present in the moments annotation json files | 0 | 12 | February 28, 2025 | |
| Could Someone Give me Advice on Integrating Ego4D Data for Object Recognition Projects? | 1 | 78 | September 5, 2024 |