Research map / Vision-language and multimodal / Human and embodied modeling
Deep Learning the City: Quantifying Urban Perception at a Global Scale
Abhimanyu Dubey, Nikhil Naik, Devi Parikh, Ramesh Raskar, César A. Hidalgo
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 on Place Pulse 2.0 with the Siamese fusion network (AlexNet, softmax) | AlexNet · Place Pulse 2.0 safe pairwise test | pairwise prediction accuracy: 53.0% -> 60.3% | Improved |
| fine-tune SS-CNN on PP 2.0 (VGGNet, RankSVM) | VGGNet · Place Pulse 2.0 safe pairwise test | pairwise prediction accuracy: 63.5% -> 72.4% | Improved |
| fine-tune SS-CNN on PP 2.0 (PlacesNet, RankSVM) | PlacesNet · Place Pulse 2.0 safe pairwise test | pairwise prediction accuracy: 61.6% -> 68.1% | Improved |
| add ranking sub-network and ranking loss to SS-CNN (VGGNet, vs softmax) (ablation) | SS-CNN (VGGNet) · Place Pulse 2.0 safe pairwise test | pairwise prediction accuracy: 65.3% -> 73.5% | Improved |
| add ranking loss to SS-CNN (AlexNet, vs softmax) (ablation) | SS-CNN (AlexNet) · Place Pulse 2.0 safe pairwise test | pairwise prediction accuracy: 60.3% -> 64.1% | Improved |
| add ranking loss to SS-CNN (PlacesNet, vs softmax) (ablation) | SS-CNN (PlacesNet) · Place Pulse 2.0 safe pairwise test | pairwise prediction accuracy: 62.2% -> 68.8% | Improved |
| use RankSVM instead of softmax output on SS-CNN (VGGNet) (ablation) | SS-CNN (VGGNet) · Place Pulse 2.0 safe pairwise test | pairwise prediction accuracy: 65.3% -> 72.4% | Improved |
| use TrueSkill instead of softmax output on SS-CNN (VGGNet) (ablation) | SS-CNN (VGGNet) · Place Pulse 2.0 safe pairwise test | pairwise prediction accuracy: 65.3% -> 67.8% | Improved |
| increase visual diversity: PP 2.0 subset instead of PP 1.0 at same data size (AlexNet) (ablation) | RSS-CNN (AlexNet) · Place Pulse 2.0 safe pairwise test | pairwise prediction accuracy: 59.8% -> 61.9% | Improved |
| increase visual diversity: PP 2.0 subset instead of PP 1.0 at same data size (VGGNet) (ablation) | RSS-CNN (VGGNet) · Place Pulse 2.0 safe pairwise test | pairwise prediction accuracy: 64.1% -> 64.2% | No clear effect |
| train on all PP 2.0 comparisons instead of PP 1.0 (ablation) | RSS-CNN · Place Pulse 2.0 safe pairwise test | pairwise prediction accuracy: 73.5% vs 64.1% for VGGNet | Improved |
| use SS-CNN (VGGNet) deep features instead of generic features (GIST, Texton, CIELab) | RankSVM · Place Pulse 2.0 safe pairwise test | pairwise prediction accuracy: 56.7% -> 73.5% | Improved |
Closest papers by meaning
- UrbanFeel:A Comprehensive Benchmark for Temporal and Perceptual Understanding of City Scenes through Human Perspective (ICLR 2026 · similarity 0.88)
- UrbanDataLayer: A Unified Data Pipeline for Urban Science (NeurIPS 2024 · similarity 0.86)
- CityLens: Evaluating Large Vision-Language Models for Urban Socioeconomic Sensing (ICLR 2026 · similarity 0.85)
- UrbanIng-V2X: A Large-Scale Multi-Vehicle, Multi-Infrastructure Dataset Across Multiple Intersections for Cooperative Perception (NeurIPS 2025 · similarity 0.85)
- Urban Region Embedding via Multi-View Contrastive Prediction (AAAI 2024 · similarity 0.85)
- Improving Region Representation Learning from Urban Imagery with Noisy Long-Caption Supervision (AAAI 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.
