Upload WiSER private archive chunk: code_snapshots

13a1c10 verified about 1 month ago

15.4 kB

	<!DOCTYPE html>
	<html lang="en">
	<head>
	<meta charset="utf-8">
	<meta name="description"
	content="WiSER is a wireless scene encoder for geometry-grounded radiomap and multipath channel impulse response prediction from sparse 3D indoor scenes.">
	<meta name="keywords"
	content="WiSER, wireless scene encoder, radiomap prediction, channel impulse response, CIR, sparse 3D voxel, Sionna, ScanNet++">
	<meta name="viewport" content="width=device-width, initial-scale=1">
	<title>WiSER: A Wireless Scene Encoder</title>

	<link href="https://fonts.googleapis.com/css?family=Google+Sans\|Noto+Sans\|Castoro"
	rel="stylesheet">
	<link rel="stylesheet" href="./static/css/bulma.min.css">
	<link rel="stylesheet" href="./static/css/fontawesome.all.min.css">
	<link rel="stylesheet"
	href="https://cdn.jsdelivr.net/gh/jpswalsh/academicons@1/css/academicons.min.css">
	<link rel="stylesheet" href="./static/css/index.css">
	<link rel="icon" href="./static/images/favicon.svg">

	<script defer src="./static/js/fontawesome.all.min.js"></script>
	<script defer src="./static/js/index.js"></script>
	</head>
	<body>

	<section class="hero wiser-hero">
	<div class="hero-body">
	<div class="container is-max-desktop">
	<div class="hero-copy has-text-centered">
	<div class="hero-quick-links">
	<a class="quick-link" href="https://scp10086.github.io/">
	<span class="icon"><i class="fas fa-home"></i></span>
	<span>Home</span>
	</a>
	<div class="quick-link dropdown is-hoverable">
	<div class="dropdown-trigger">
	<button class="button is-white is-rounded research-button" aria-haspopup="true" aria-controls="research-menu">
	<span>More Research</span>
	<span class="icon is-small"><i class="fas fa-angle-down" aria-hidden="true"></i></span>
	</button>
	</div>
	<div class="dropdown-menu" id="research-menu" role="menu">
	<div class="dropdown-content">
	<a class="dropdown-item" href="https://scp10086.github.io/wiser-page/">WiSER</a>
	<span class="dropdown-item is-muted">More projects coming soon</span>
	</div>
	</div>
	</div>
	</div>
	<p class="eyebrow">Geometry-Grounded Wireless Prediction</p>
	<h1 class="title publication-title">
	WiSER: A Wireless Scene Encoder for Geometry-Grounded Multi-View Wireless Prediction
	</h1>
	<p class="subtitle wiser-subtitle">
	A sparse 3D scene representation that jointly supports dense
	radiomap prediction and multipath channel impulse response (CIR)
	tap-set prediction.
	</p>

	<div class="publication-authors">
	<span class="author-block">
	<a href="https://scp10086.github.io/">Jing Qiao</a></span>,
	<span class="author-block">
	<a href="#" class="author-link-placeholder" title="Personal page coming soon">Yiyang Guo</a></span>,
	<span class="author-block">
	<a href="#" class="author-link-placeholder" title="Personal page coming soon">Hao Ye</a></span>
	</div>
	<div class="publication-authors institution">
	University of California Santa Cruz
	</div>

	<div class="publication-links">
	<span class="link-block">
	<span class="button is-normal is-rounded is-dark is-static-like">
	<span class="icon"><i class="fas fa-file-pdf"></i></span>
	<span>Paper coming soon</span>
	</span>
	</span>
	<span class="link-block">
	<span class="button is-normal is-rounded is-dark is-static-like">
	<span class="icon"><i class="fab fa-github"></i></span>
	<span>Code coming soon</span>
	</span>
	</span>
	<span class="link-block">
	<span class="button is-normal is-rounded is-dark is-static-like">
	<span class="icon"><i class="fas fa-database"></i></span>
	<span>Data coming soon</span>
	</span>
	</span>
	<span class="link-block">
	<span class="button is-normal is-rounded is-dark is-static-like">
	<span class="icon"><i class="fas fa-cube"></i></span>
	<span>Model coming soon</span>
	</span>
	</span>
	</div>
	</div>
	</div>
	</div>
	</section>

	<section id="abstract" class="section">
	<div class="container is-max-desktop">
	<div class="columns is-centered has-text-centered">
	<div class="column is-four-fifths">
	<h2 class="title is-3">Abstract</h2>
	<div class="content has-text-justified">
	<p>
	Indoor wireless propagation is governed by the interaction among
	3D scene geometry, radio-material properties, and transmitter and
	receiver configuration. Most learning-based site-specific
	prediction methods focus on a single wireless representation, such
	as radiomap estimation or CIR prediction, and therefore do not
	explicitly exploit the propagation structure shared across
	heterogeneous wireless views.
	</p>
	<p>
	WiSER maps a sparse voxel representation of an indoor scene and a
	transmitter location into a transmitter-conditioned sparse 3D scene
	memory. This shared memory is queried by two structure-aware
	decoders: a ray-corridor decoder for dense receiver-plane path-gain
	prediction and a Detection Transformer-style set decoder for
	variable-cardinality delay and power tap prediction.
	</p>
	<p>
	We train and evaluate WiSER with a co-registered indoor
	scene--wireless dataset generated from ScanNet++ scenes and Sionna
	Ray Tracing. The dataset aligns sparse voxel inputs, dense radiomap
	labels, and unordered multipath CIR tap sets under a common
	coordinate frame and propagation configuration.
	</p>
	</div>
	</div>
	</div>
	</div>
	</section>

	<section class="section is-light">
	<div class="container is-max-widescreen">
	<h2 class="title is-3 has-text-centered">Key Idea</h2>
	<div class="columns is-variable is-5">
	<div class="column">
	<div class="feature-card">
	<div class="feature-index">01</div>
	<h3 class="title is-5">Shared Wireless Scene Memory</h3>
	<p>
	WiSER encodes sparse 3D scene voxels into a transmitter-conditioned
	memory that can be reused by multiple wireless prediction views.
	</p>
	</div>
	</div>
	<div class="column">
	<div class="feature-card">
	<div class="feature-index">02</div>
	<h3 class="title is-5">Ray-Corridor Radiomap Decoding</h3>
	<p>
	The radiomap branch gathers receiver-specific scene tokens near the
	transmitter--receiver corridor to decode dense path-gain fields.
	</p>
	</div>
	</div>
	<div class="column">
	<div class="feature-card">
	<div class="feature-index">03</div>
	<h3 class="title is-5">DETR-Style CIR Set Prediction</h3>
	<p>
	The CIR branch predicts unordered multipath delay--power taps with
	learnable path queries and Hungarian matching.
	</p>
	</div>
	</div>
	</div>
	</div>
	</section>

	<section id="method" class="section">
	<div class="container is-max-widescreen">
	<h2 class="title is-3 has-text-centered">Method Overview</h2>
	<div class="figure-panel">
	<img src="./static/images/wiser/architecture.png"
	alt="Overall WiSER architecture">
	<p class="caption">
	WiSER first builds a transmitter-conditioned sparse 3D scene memory.
	A ray-corridor radiomap decoder predicts dense receiver-plane path gain,
	while a CIR set decoder predicts variable-cardinality delay and power
	taps for a specific transmitter--receiver link.
	</p>
	</div>

	<div class="columns is-variable is-6 method-columns">
	<div class="column is-half">
	<h3 class="title is-4">Ray-corridor feature gathering</h3>
	<p>
	For each receiver query, WiSER selects a compact set of scene voxels
	near the transmitter--receiver segment and endpoint neighborhoods.
	This gives the radiomap decoder access to likely blockers,
	openings, reflectors, and nearby scattering structures without dense
	attention over the full indoor volume.
	</p>
	</div>
	<div class="column is-half">
	<div class="figure-panel compact">
	<img src="./static/images/wiser/ray_corridor.png"
	alt="Ray-corridor feature gathering">
	</div>
	</div>
	</div>
	</div>
	</section>

	<section id="dataset" class="section is-light">
	<div class="container is-max-widescreen">
	<h2 class="title is-3 has-text-centered">Co-Registered Scene--Wireless Dataset</h2>
	<div class="columns is-variable is-6 is-vcentered">
	<div class="column is-5">
	<div class="content">
	<p>
	WiSER is trained with a co-registered dataset pipeline that converts
	indoor 3D scenes into both sparse voxel inputs for learning and
	Sionna-compatible radio scenes for ray-tracing supervision.
	</p>
	<p>
	The same coordinate frame produces aligned dense radiomap labels
	and path-level CIR labels. This makes it possible to study a single
	learned scene representation across coverage-level and path-level
	wireless views.
	</p>
	<div class="metric-grid">
	<div class="metric-card">
	<span class="metric-value">100</span>
	<span class="metric-label">training scenes</span>
	</div>
	<div class="metric-card">
	<span class="metric-value">10 cm</span>
	<span class="metric-label">voxel size</span>
	</div>
	<div class="metric-card">
	<span class="metric-value">2</span>
	<span class="metric-label">wireless views</span>
	</div>
	</div>
	</div>
	</div>
	<div class="column is-7">
	<div class="figure-panel">
	<img src="./static/images/wiser/dataset_pipeline.png"
	alt="WiSER dataset generation pipeline">
	<p class="caption">
	The dataset pipeline aligns sparse voxel scenes, Sionna material
	scenes, dense radiomaps, and multipath CIR tap sets.
	</p>
	</div>
	</div>
	</div>
	</div>
	</section>

	<section id="results" class="section">
	<div class="container is-max-widescreen">
	<h2 class="title is-3 has-text-centered">Results</h2>
	<div class="columns is-variable is-5">
	<div class="column">
	<div class="result-card">
	<h3 class="title is-5">Radiomap Prediction</h3>
	<div class="result-number">3.834 dB</div>
	<p class="result-label">MAE on evaluated radiomap cases</p>
	<p>
	WiSER improves over scene-specific NeRF2 and RF-3DGS baselines
	while being trained once across multiple scenes.
	</p>
	</div>
	</div>
	<div class="column">
	<div class="result-card">
	<h3 class="title is-5">Multipath CIR Prediction</h3>
	<div class="result-number">5.89 dB</div>
	<p class="result-label">matched peak-power MAE</p>
	<p>
	The CIR decoder reduces matched peak-power and delay errors over
	geometry-only and 3D CNN reference baselines.
	</p>
	</div>
	</div>
	<div class="column">
	<div class="result-card">
	<h3 class="title is-5">Shared Representation</h3>
	<div class="result-number">0.61 ns</div>
	<p class="result-label">matched delay MAE</p>
	<p>
	Results support the central claim that a sparse 3D scene memory can
	serve both dense field-level and sparse path-level prediction.
	</p>
	</div>
	</div>
	</div>

	<div class="columns is-variable is-6">
	<div class="column is-half">
	<div class="figure-panel">
	<img src="./static/images/wiser/radiomap_qualitative.png"
	alt="Qualitative radiomap comparison">
	<p class="caption">
	Qualitative radiomap examples compare ground truth, WiSER, NeRF2,
	RF-3DGS, and radiomap-head ablations under the same dB color scale.
	</p>
	</div>
	</div>
	<div class="column is-half">
	<div class="figure-panel">
	<img src="./static/images/wiser/cir_qualitative.png"
	alt="Qualitative CIR prediction comparison">
	<p class="caption">
	Qualitative CIR examples show matched predicted taps against
	ground-truth delay--power taps in the delay/power plane.
	</p>
	</div>
	</div>
	</div>

	<div class="table-container wiser-table">
	<table class="table is-fullwidth is-hoverable">
	<thead>
	<tr>
	<th>Task</th>
	<th>Method</th>
	<th>Primary Metric</th>
	<th>Additional Metrics</th>
	</tr>
	</thead>
	<tbody>
	<tr>
	<td>Radiomap</td>
	<td><strong>WiSER</strong></td>
	<td><strong>3.834 dB MAE</strong></td>
	<td>5.500 dB RMSE, 26.78 dB PSNR</td>
	</tr>
	<tr>
	<td>Radiomap</td>
	<td>RF-3DGS</td>
	<td>4.585 dB MAE</td>
	<td>6.281 dB RMSE, 25.62 dB PSNR</td>
	</tr>
	<tr>
	<td>CIR</td>
	<td><strong>WiSER</strong></td>
	<td><strong>5.89 dB peak-power MAE</strong></td>
	<td>0.61 ns delay MAE, 0.477 count accuracy</td>
	</tr>
	<tr>
	<td>CIR</td>
	<td>3D CNN</td>
	<td>11.50 dB peak-power MAE</td>
	<td>1.50 ns delay MAE, 0.407 count accuracy</td>
	</tr>
	</tbody>
	</table>
	</div>
	</div>
	</section>

	<section id="release" class="section is-light">
	<div class="container is-max-desktop">
	<h2 class="title is-3 has-text-centered">Release Status</h2>
	<div class="release-panel">
	<p>
	We are preparing the public release of the WiSER codebase, processed
	dataset, and model checkpoint. The current project page is a preview;
	public links will be added after the corresponding repositories and
	archives are finalized.
	</p>
	<div class="release-list">
	<span>Code: coming soon</span>
	<span>Dataset: coming soon</span>
	<span>Model: coming soon</span>
	<span>Paper/arXiv: coming soon</span>
	</div>
	</div>
	</div>
	</section>

	<footer class="footer">
	<div class="container">
	<div class="content has-text-centered">
	<p>
	This project page is adapted from a public academic project-page template.
	</p>
	<p>
	WiSER project page, 2026.
	</p>
	</div>
	</div>
	</footer>

	</body>
	</html>