Monocular Building Height Estimation from PhiSat-2 Imagery: Dataset and Method
arXiv:2603.29245v1 Announce Type: new Abstract: Monocular building height estimation from optical imagery is important for urban morphology characterization but remains challenging due to ambiguous height cues, large inter-city variations in building morphology, and the long-tailed distribution of building heights. PhiSat-2 is a promising open-access data source for this task because of its global coverage, 4.75 m spatial resolution, and seven-band spectral observations, yet its potential has not been systematically evaluated. To address this gap, we construct a PhiSat-2-Height dataset (PHDataset) and propose a Two-Stream Ordinal Network (TSONet). PHDataset contains 9,475 co-registered image-label patch pairs from 26 cities worldwide. TSONet jointly models footprint segmentation and height
View PDF HTML (experimental)
Abstract:Monocular building height estimation from optical imagery is important for urban morphology characterization but remains challenging due to ambiguous height cues, large inter-city variations in building morphology, and the long-tailed distribution of building heights. PhiSat-2 is a promising open-access data source for this task because of its global coverage, 4.75 m spatial resolution, and seven-band spectral observations, yet its potential has not been systematically evaluated. To address this gap, we construct a PhiSat-2-Height dataset (PHDataset) and propose a Two-Stream Ordinal Network (TSONet). PHDataset contains 9,475 co-registered image-label patch pairs from 26 cities worldwide. TSONet jointly models footprint segmentation and height estimation, and introduces a Cross-Stream Exchange Module (CSEM) and a Feature-Enhanced Bin Refinement (FEBR) module for footprint-aware feature interaction and ordinal height refinement. Experiments on PHDataset show that TSONet achieves the best overall performance, reducing MAE and RMSE by 13.2% and 9.7%, and improving IoU and F1-score by 14.0% and 10.1% over the strongest competing results. Ablation studies further verify the effectiveness of CSEM, FEBR, and the joint use of ordinal regression and footprint assistance. Additional analyses indicate that PhiSat-2 benefits monocular building height estimation through its balanced combination of building-relevant spatial detail and multispectral observations. Overall, this study confirms the potential of PhiSat-2 for monocular building height estimation and provides a dedicated dataset and an effective method for future research.
Subjects:
Computer Vision and Pattern Recognition (cs.CV)
Cite as: arXiv:2603.29245 [cs.CV]
(or arXiv:2603.29245v1 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2603.29245
arXiv-issued DOI via DataCite (pending registration)
Submission history
From: Yanjiao Song [view email] [v1] Tue, 31 Mar 2026 04:14:57 UTC (32,674 KB)
Sign in to highlight and annotate this article
Conversation starters
Daily AI Digest
Get the top 5 AI stories delivered to your inbox every morning.
More about
modelannouncefeature
Blazor WASM's Deputy Thread Model Will Break JavaScript Interop - Here's Why That Matters
<h2> The Problem </h2> <p>Microsoft is changing how .NET runs inside WebAssembly. When you enable threading with <code><WasmEnableThreads>true</WasmEnableThreads></code>, the entire .NET runtime moves off the browser's main thread and onto a background Web Worker — what they call the <strong>"Deputy Thread" model</strong>.</p> <p>This sounds like a good idea on paper. The UI stays responsive. .NET gets real threads. Everyone wins.</p> <p>Except it breaks JavaScript interop. Not in a subtle, edge-case way. It breaks it <em>fundamentally</em>.</p> <h2> What Actually Happens </h2> <p>In traditional Blazor WASM (no threading), .NET and JavaScript share the same thread. When JavaScript calls <code>DotNet.invokeMethod</code>, the CPU jumps from the JS stack to the C# stack and back. It's fast. I
The Wrong Way to Use AI for Debugging (And the Mental Model That Actually Works)
<p>Same AI tool. Same codebase access. Same staging incident. Three experienced engineers spent hours and couldn't find the root cause. I found it in about 20 minutes — and I barely knew what half the tools did.</p> <p>This isn't a story about juniors being better than seniors. It's about a mental model for AI collaboration that anyone can use — and why the "obvious" way most people use AI actually holds them back.</p> <h2> The Incident </h2> <p>A message hits the team Slack channel:</p> <blockquote> <p>"Staging is broken. We're investigating."</p> </blockquote> <p>The error:<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code>alembic current ERROR: Can't locate revision identified by 'ba29cdc8739d' FAILED: Can't locate revision identified by 'ba29cdc8
The hidden cost of GPT-4o: what every SaaS founder should know about per-user LLM spend it
<p>So you're running a SaaS that leans on an LLM. You check your OpenAI bill at the end of the month, it's a few hundred bucks, you shrug and move on. As long as it's not five figures, who cares, right?</p> <p>Wrong. That total is hiding a nasty secret: you're probably losing money on some of your users.</p> <p>I'm not talking about the obvious free-tier leeches. I'm talking about paying customers who are costing you more in API calls than they're giving you in subscription fees. You're literally paying for them to use your product.</p> <p><strong>The problem with averages</strong></p> <p>Let's do some quick, dirty math. GPT-4o pricing settled at around $3/1M tokens for input and $10/1M for output. It's cheap, but it's not free.</p> <p>Say you have a summarization feature. A user pastes in
Knowledge Map
Connected Articles — Knowledge Graph
This article is connected to other articles through shared AI topics and tags.
More in Research Papers
Illinois Tech computer science researcher honored by IEEE Chicago Section - EurekAlert!
<a href="https://news.google.com/rss/articles/CBMiXEFVX3lxTE13OVpWMEk1Z3hlMkR2bHNBQ2dkazFwb3VqN3hCa29GWGJvSVlPa00zd2xUakRmYXFqQmc5OWU0eGl4a21FMDAwWUN2Q3p0M3FrbXBkNV8zN0cxaG1s?oc=5" target="_blank">Illinois Tech computer science researcher honored by IEEE Chicago Section</a> <font color="#6f6f6f">EurekAlert!</font>
AI maps science papers to predict research trends two to three years ahead - Tech Xplore
<a href="https://news.google.com/rss/articles/CBMie0FVX3lxTE5aTkZYTWdaRDZwTXNRMldpMG1WZ1YzWDZTOHN5M183Z3A1ZTFYbnhEWTdPRmpvZnZFU0xodlRsNWxFaGxTcEpwalhJNmJpQWE5VjhaRS1tOXJIeTc5Z0JNblJ3dFd4WjRYZGJOX0NrWGt6ZmZJVTBpRm5wWQ?oc=5" target="_blank">AI maps science papers to predict research trends two to three years ahead</a> <font color="#6f6f6f">Tech Xplore</font>
AI inspires new research topics in materials science - Nanowerk
<a href="https://news.google.com/rss/articles/CBMiZ0FVX3lxTFBPWlJSM2ExeVQ3LVppTm45NHpEMW9YVkxscThCNDd2OVB0c3J1ZmVCbWNSZWZ0TjZwSzlOdEFXN2UtRk5LU1hxdXd4ZklldGxoM0FZSnhCd19PWkNHQ1ZRVDNwSHNUSk0?oc=5" target="_blank">AI inspires new research topics in materials science</a> <font color="#6f6f6f">Nanowerk</font>
Discussion
Sign in to join the discussion
No comments yet — be the first to share your thoughts!