Research map / Vision-language and multimodal / Visual question answering
EgoNight: Towards Egocentric Vision Understanding at Night with a Challenging Benchmark
Deheng Zhang, Yuqian Fu, Runyi Yang, Yang Miao, Tianwen Qian, Xu Zheng, Guolei Sun, Ajad Chhatkuli, Xuanjing Huang, Yu-Gang Jiang, Luc Van Gool, Danda Pani Paudel
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 |
|---|---|---|---|
| evaluate nighttime vs daytime for egocentric MLLMs | MLLMs (averaged) · EgoNight-Synthetic | VQA accuracy: -32.8% | Hurt |
| scale Qwen2.5-VL from 7B to 72B | Qwen2.5-VL · EgoNight-VQA avg | accuracy (%): 13.44 -> 18.99 | Improved |
| full fine-tuning on EgoNight train split vs zero-shot (ablation) | Qwen2.5-VL-7B · EgoNight real | accuracy (%): 16.40 -> 25.61 | Improved |
| tune LLM only vs vision encoder only (ablation) | Qwen2.5-VL-7B · EgoNight synthetic | accuracy (%): 29.74 -> 35.50 | Improved |
| full fine-tuning vs LLM-only tuning (ablation) | Qwen2.5-VL-7B · EgoNight synthetic | accuracy (%): 35.50 -> 36.25 | Improved |
| fine-tune on synthetic nighttime data only (ablation) | Qwen2.5-VL-7B · EgoNight real | accuracy (%): 14.83 -> 20.57 | Improved |
| fine-tune (LLM-only) on reasoning tasks (navigation) (ablation) | Qwen2.5-VL-7B · EgoNight navigation | accuracy (%): 17.870 -> 19.918 | Improved |
| fine-tune (LLM-only) on perception tasks (object recognition) (ablation) | Qwen2.5-VL-7B · EgoNight object recognition | accuracy (%): 8.435 -> 35.855 | Improved |
| use night query instead of day query for retrieval | GPT-4.1 · EgoNight-Synthetic spatial retrieval | Acc R@1 (%): 75.6 -> 54.1 | Hurt |
| MLLM (GPT-4.1) vs feature-based retrieval for temporal localization | GPT-4.1 · EgoNight-Sofia Night->Day | mIoU (%): 33.4 -> 15.5 | Hurt |
| use fisheye-specific depth model (UniK3D) vs general Depth Anything | depth estimation · EgoNight-Synthetic night | Abs Rel: 0.302 -> 0.253 | Improved |
Closest papers by meaning
- Seeing in the Dark: Benchmarking Egocentric 3D Vision with the Oxford Day-and-Night Dataset (NeurIPS 2025 · similarity 0.88)
- Omnia de EgoTempo: Benchmarking Temporal Understanding of Multi-Modal LLMs in Egocentric Videos (CVPR 2025 · similarity 0.87)
- EgoBlind: Towards Egocentric Visual Assistance for the Blind (NeurIPS 2025 · similarity 0.87)
- EgoTextVQA: Towards Egocentric Scene-Text Aware Video Question Answering (CVPR 2025 · similarity 0.86)
- ANNEXE: Unified Analyzing, Answering, and Pixel Grounding for Egocentric Interaction (CVPR 2025 · similarity 0.85)
- EgoLife: Towards Egocentric Life Assistant (CVPR 2025 · similarity 0.85)
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.
