Research map / Video / Video-language understanding
CARIM: Caption-Based Autonomous Driving Scene Retrieval via Inclusive Text Matching
Minjoo Ki, Daejung Kim, Kisung Kim, Seon Joo Kim, Jinhan Lee
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 |
|---|---|---|---|
| add Adaptive Negative Injection to Balanced Sampling (ablation) | CARIM negative query generation · DRAMA test | F1-score: 52.88 -> 73.5 | Improved |
| add Balanced Sampling to Adaptive Negative Injection (ablation) | CARIM negative query generation · DRAMA test | F1-score: 70.26 -> 73.5 | Improved |
| combine all three losses (pos, neg, self-alignment) instead of single loss (ablation) | CARIM · DRAMA test | F1-score: 8.3 -> 73.5 | Improved |
| train with negative loss only (ablation) | CARIM · DRAMA test | F1-score: 73.5 -> 0.0 | Hurt |
| replace ranking-based video retrieval (DRL) with caption-based text matching (CARIM) | driving scene retrieval · DRAMA test | inference time (s): 257.1 -> 3.58 | Improved |
| use object-level attention scoring instead of whole-caption SBERT similarity | CARIM vs SBERT-RB · DRAMA test | negative accuracy: 43.9 -> 99.3 | Improved |
| object-level attention scoring vs SBERT-RB on 4-object queries, similar recall | CARIM vs SBERT-RB · DRAMA test 4 objects | Similar-query recall: 62.5 -> 56.2 | Hurt |
| object-level scoring on single-object queries | CARIM vs SBERT-RB · DRAMA test 1 object | F1-score: 1.3 -> 62.9 | Improved |
Closest papers by meaning
- BEV-TSR: Text-Scene Retrieval in BEV Space for Autonomous Driving (AAAI 2025 · similarity 0.89)
- Driving Video Retrieval for Complex Queries with Structured Grounding (NeurIPS 2026 · similarity 0.89)
- MAPLM: A Real-World Large-Scale Vision-Language Benchmark for Map and Traffic Scene Understanding (CVPR 2024 · similarity 0.83)
- DriveSpatial: A Benchmark for Spatiotemporal Intelligence in VLMs for Autonomous Driving (NeurIPS 2026 · similarity 0.83)
- Prune2Drive: A Plug-and-Play Framework for Accelerating Vision-Language Models in Autonomous Driving (CVPR 2026 · similarity 0.83)
- DriveDreamer-2: LLM-Enhanced World Models for Diverse Driving Video Generation (AAAI 2025 · similarity 0.83)
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.
