Monocular Building Height Estimation from PhiSat-2 Imagery: Dataset and Method

2026-03-31 · Source: Takara TLDR - Daily AI Papers · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Expert, medium

Summary

Researchers have developed a new dataset and method for monocular building height estimation using PhiSat-2 satellite imagery. The PhiSat-2-Height dataset (PHDataset) comprises 9,475 co-registered image-label patch pairs from 26 global cities, leveraging PhiSat-2's 4.75 m spatial resolution and seven-band spectral observations. They also propose a Two-Stream Ordinal Network (TSONet) that integrates footprint segmentation and height estimation. TSONet incorporates a Cross-Stream Exchange Module (CSEM) for footprint-aware feature interaction and a Feature-Enhanced Bin Refinement (FEBR) module for ordinal height refinement. Experiments on PHDataset show TSONet reduces MAE and RMSE by 13.2% and 9.7% respectively, and improves IoU and F1-score by 14.0% and 10.1% compared to leading competitors, confirming PhiSat-2's utility for this task.

Key takeaway

For research scientists working on urban morphology or 3D city modeling, this work demonstrates that PhiSat-2 imagery offers a robust, open-access data source for monocular building height estimation. You should consider integrating the PHDataset and TSONet's joint modeling approach into your workflows to achieve higher accuracy in building height predictions, especially when dealing with diverse urban landscapes and ambiguous height cues.

Key insights

PhiSat-2 imagery, combined with a novel two-stream ordinal network, significantly improves monocular building height estimation.

Principles

• Joint modeling improves performance.
• Ordinal regression refines height estimation.
• Multispectral data enhances spatial detail.

Method

TSONet jointly models footprint segmentation and height estimation using a Cross-Stream Exchange Module for feature interaction and a Feature-Enhanced Bin Refinement module for ordinal height refinement.

In practice

• Utilize PhiSat-2 for urban morphology tasks.
• Employ joint segmentation and height models.
• Incorporate ordinal regression for height.

Topics

• Monocular Building Height Estimation
• PhiSat-2 Imagery
• PHDataset
• Two-Stream Ordinal Network
• Footprint Segmentation

Code references

• AIGeeksGroup/AnyDepth
• farooqhassaan/FlyPose
• wuaodi/SpaceSense-Bench

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, Computer Vision Engineer

Related on AIssential

• PCFootprint: A Large-Scale Dataset and Benchmark for Vectorized Building Footprint Extraction from Aerial LiDAR Point Clouds — PCFootprint is the first large-scale LiDAR dataset for vectorized building footprint extraction, addressing optical imagery limitations.
• SyMTRS: Benchmark Multi-Task Synthetic Dataset for Depth, Domain Adaptation and Super-Resolution in Aerial Imagery — SyMTRS is a multi-task synthetic dataset for aerial imagery, providing perfect ground truth for depth, domain adaptation, and super-resolution.
• LatticeVision: Image to Image Networks for Modeling Non-Stationary Spatial Data — Image-to-image networks globally estimate non-stationary spatial parameters faster and more accurately than local methods.
• Honey, I Shrunk the Arc de Triomphe! — A new dataset and method mitigate scale-collapse in monocular depth estimation for distant, unconstrained scenes.
• We’re announcing Canopy Height Maps v2 (CHMv2), an open source model for high-resolution global forest canopy mapping, developed in partnership with the @WorldResources. CHMv2 leverages our DINOv3 Sat-L vision model, specifically optimized for s - x.com — CHMv2, an open-source model, uses DINOv3 Sat-L for high-resolution global forest canopy mapping.
• Zero-Shot Polygon Matching with Pre-trained Models for Pose Estimation and Polygon Cloud from Challenging Stereo — Unsupervised polygon matching for stereo images is achievable by integrating pre-trained models with handcrafted features.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Takara TLDR - Daily AI Papers.