Research map / Video / Video segmentation and tracking
STREAMER: Streaming Representation Learning and Event Segmentation in a Hierarchical Manner
Ramy Mounir, Sujal Vijayaraghavan, Sudeep Sarkar
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use STREAMER instead of Offline ABD | unsupervised event segmentation · EPIC-KITCHENS Protocol 1 | MoF: 0.704 -> 0.759 | Improved |
| use STREAMER instead of Online ABD | unsupervised event segmentation · EPIC-KITCHENS Protocol 1 | IoU: 0.496 -> 0.508 | Improved |
| use full 3-layer hierarchy (HLR) instead of best single layer | STREAMER · EPIC-KITCHENS Protocol 1 | MoF: 0.759 -> 0.736 | Hurt |
| use full 3-layer hierarchy (HLR) instead of best single layer | STREAMER · EPIC-KITCHENS Protocol 1 | IoU: 0.508 -> 0.511 | Improved |
| evaluate on unseen kitchens (Protocol 1) vs seen kitchens (Protocol 2) | STREAMER · EPIC-KITCHENS Protocol 1 | MoF: 0.759 -> 0.754 | No clear effect |
| self-supervised STREAMER features instead of supervised pretrained features | event retrieval · EPIC-KITCHENS | MSE: 1.00 -> 0.967 | Improved |
| swap temporal encoder from GRU to LSTM (ablation) | STREAMER · EPIC-KITCHENS Protocol 1 (best) | MoF: 0.759 -> 0.761 | No clear effect |
| swap temporal encoder from GRU to Transformer (ablation) | STREAMER · EPIC-KITCHENS Protocol 1 (best) | MoF: 0.759 -> 0.759 | No clear effect |
| replace average predictor with layer-specific MLP on the average (ablation) | STREAMER predictor p · EPIC-KITCHENS Protocol 1 (best) | MoF: 0.742 -> 0.749 | Improved |
| replace average + MLP predictor with transformer retrieving cross-layer context (ablation) | STREAMER predictor p · EPIC-KITCHENS Protocol 1 (best) | MoF: 0.749 -> 0.759 | Improved |
| reach of influence alpha 1 to 2 (ablation) | STREAMER · EPIC-KITCHENS Protocol 1 (best) | MoF: 0.756 -> 0.797 | Improved |
| reach of influence alpha 2 to 3 (ablation) | STREAMER · EPIC-KITCHENS Protocol 1 (best) | MoF: 0.797 -> 0.759 | Hurt |
Closest papers by meaning
- Learning Streaming Video Representation via Multitask Training (ICCV 2025 · similarity 0.88)
- Actor-Centered Representations for Action Localization in Streaming Videos (ECCV 2022 · similarity 0.87)
- StreamForest: Efficient Online Video Understanding with Persistent Event Memory (NeurIPS 2025 · similarity 0.85)
- Learning from Streaming Video with Orthogonal Gradients (CVPR 2025 · similarity 0.85)
- Efficient Event-Based Semantic Segmentation via Exploiting Frame-Event Fusion: A Hybrid Neural Network Approach (AAAI 2025 · similarity 0.85)
- Streaming Long Video Understanding with Large Language Models (NeurIPS 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
