mlsanityData to 14 Oct 2026

Storing Less, Finding More: How Novelty Filtering Improves Cross-Modal Retrieval on Edge Cameras

arXiv 2603.29631 ↗First posted 31 Mar 2026cs.CV cs.DC cs.IRcited by 0

Not on the map. Why?

Abstract (arXiv)

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.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 18 of its 21 references that are in the map. First posted on arXiv 31 Mar 2026.

Builds on

18 of 21 references.

References · 18

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  2. 2023Sigmoid Loss for Language Image Pre-Trainingcited by 3,109
  3. 2021Ego4D: Around the World in 3,000 Hours of Egocentric Videocited by 1,556
  4. 2025SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Featurescited by 1,188
  5. 2023ImageBind: One Embedding Space To Bind Them Allcited by 1,175
  6. 2021Deduplicating Training Data Makes Language Models Bettercited by 714
  7. 2021LiT: Zero-Shot Transfer with Locked-image text Tuningcited by 577
  8. 2020Rescaling Egocentric Visioncited by 540
  9. 2021Less is More: ClipBERT for Video-and-Language Learning via Sparse Samplingcited by 484
  10. 2023SemDeDup: Data-efficient learning at web-scale through semantic deduplicationcited by 303
  11. 2020TransNet V2: An effective deep network architecture for fast shot transition detectioncited by 171
  12. 2023EgoVLPv2: Egocentric Video-Language Pre-training with Fusion in the Backbonecited by 130
6 more
  1. 2021CLIP-It! Language-Guided Video Summarizationcited by 99
  2. 2023MobileCLIP: Fast Image-Text Models through Multi-Modal Reinforced Trainingcited by 96
  3. 2023TinyCLIP: CLIP Distillation via Affinity Mimicking and Weight Inheritancecited by 90
  4. 2024KeyVideoLLM: Towards Large-scale Video Keyframe Selectioncited by 54
  5. 2025Empowering Edge Intelligence: A Comprehensive Survey on On-Device AI Modelscited by 46
  6. 2025Full System Architecture Modeling for Wearable Egocentric Contextual AIcited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026DREAM: Extending Vision-Language Models with Dual-Objective Encoding for Cross-Modal Retrieval
  2. 2025Cross the Gap: Exposing the Intra-modal Misalignment in CLIP via Modality Inversion
  3. 2026A Strong Baseline for Evaluating Vision Encoders in Multimodal Large Language Models
  4. 2025Divide, then Ground: Adapting Frame Selection to Query Types for Long-Form Video Understanding
  5. 2025BitMar: Low-Bit Multimodal Fusion with Episodic Memory for Edge Devices
  6. 2025Venus: An Efficient Edge Memory-and-Retrieval System for VLM-based Online Video Understanding
  7. 2026OmniRetriever: Any-to-Any Audio-Video-Text Retrieval via Fusion-as-Teacher Distillation
  8. 2023Robust Cross-Modal Knowledge Distillation for Unconstrained Videos