Research map / Detection and segmentation / Open-set and few-shot detection
MINGLE: VLMs for Semantically Complex Region Detection in Urban Scenes
Liu Liu, Alexandra Schild, Marco Cipriano, Fatimeh Al Ghannam, Freya Tan, Gerard de Melo, Andres Sevtsuk
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 |
|---|---|---|---|
| fine-tune Qwen2-VL-7B on pairwise social labels instead of zero-shot | Qwen2-VL-7B · Pairwise test set | F1 (Yes): 0.53 -> 0.72 | Improved |
| use balanced data and drop Not Sure labels (ablation) | Qwen2-VL-7B LoRA 64 · Pairwise test set | F1 (Yes): 0.68 -> 0.72 | Improved |
| drop Not Sure labels on unbalanced data (ablation) | Qwen2-VL-7B LoRA 64 · Pairwise test set | F1 (Yes): 0.70 -> 0.68 | Hurt |
| increase data from 29k to 76k (ablation) | Qwen2-VL-7B LoRA 64 · Pairwise test set | F1 (Yes): 0.64 -> 0.70 | Improved |
| LoRA rank 128 instead of 64 on Qwen2.5-VL-7B (ablation) | Qwen2.5-VL-7B · Pairwise test set | F1 (No): 0.53 -> 0.39 | Hurt |
| use MINGLE pipeline vs zero-shot VLMs | MINGLE · ELSA | mIoU: 0.02 -> 0.64 | Improved |
| add distance and depth filtering of pairs (ablation) | MINGLE · ELSA | mIoU: 0.635 -> 0.634 | No clear effect |
| add distance filtering only (ablation) | MINGLE · ELSA | F1: 0.604 -> 0.608 | No clear effect |
Closest papers by meaning
- Learning Human-Object Interaction as Groups (NeurIPS 2025 · similarity 0.87)
- Beyond F-Formations: Determining Social Involvement in Free Standing Conversing Groups From Static Images (CVPR 2016 · similarity 0.87)
- Part-Aware Bottom-Up Group Reasoning for Fine-Grained Social Interaction Detection (NeurIPS 2025 · similarity 0.85)
- Eyes on VLM: Benchmarking Gaze Following and Social Gaze Prediction in Vision Language Models (NeurIPS 2026 · similarity 0.84)
- Spot The Ball: A Benchmark for Visual Social Inference (CVPR 2026 · similarity 0.84)
- UrbanFeel:A Comprehensive Benchmark for Temporal and Perceptual Understanding of City Scenes through Human Perspective (ICLR 2026 · 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.
