Saved in:
| Main Author: | |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2603.29631 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866912992091701248 |
|---|---|
| author | Abdelwahab, Sherif |
| author_facet | Abdelwahab, Sherif |
| contents | Always-on edge cameras generate continuous video streams where redundant frames degrade cross-modal retrieval by crowding correct results out of top-k search. This paper presents a streaming retrieval architecture: an on-device epsilon-net filter retains only semantically novel frames, building a denoised embedding index; a cross-modal adapter and cloud re-ranker compensate for the compact encoder's weak alignment. A single-pass streaming filter outperforms offline alternatives (k-means, farthest-point, uniform, random) across eight vision-language models (8M-632M) on two egocentric datasets (AEA, EPIC-KITCHENS). Combined, the architecture reaches 45.6% Hit@5 on held-out data using an 8M on-device encoder at an estimated 2.7 mW. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_29631 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Storing Less, Finding More: How Novelty Filtering Improves Cross-Modal Retrieval on Edge Cameras Abdelwahab, Sherif Computer Vision and Pattern Recognition Distributed, Parallel, and Cluster Computing Information Retrieval I.4.9; I.2.10 Always-on edge cameras generate continuous video streams where redundant frames degrade cross-modal retrieval by crowding correct results out of top-k search. This paper presents a streaming retrieval architecture: an on-device epsilon-net filter retains only semantically novel frames, building a denoised embedding index; a cross-modal adapter and cloud re-ranker compensate for the compact encoder's weak alignment. A single-pass streaming filter outperforms offline alternatives (k-means, farthest-point, uniform, random) across eight vision-language models (8M-632M) on two egocentric datasets (AEA, EPIC-KITCHENS). Combined, the architecture reaches 45.6% Hit@5 on held-out data using an 8M on-device encoder at an estimated 2.7 mW. |
| title | Storing Less, Finding More: How Novelty Filtering Improves Cross-Modal Retrieval on Edge Cameras |
| topic | Computer Vision and Pattern Recognition Distributed, Parallel, and Cluster Computing Information Retrieval I.4.9; I.2.10 |
| url | https://arxiv.org/abs/2603.29631 |