Structured3D is live · free · last verified 24 Oct 2026 · viability monitor
Structured3D
Free synthetic indoor-scene dataset with photo-realistic renders and dense 3D structure annotations for layout research.
If room layout and structural geometry are your target, Structured3D is one of the few free datasets that hands you dense line, plane, junction and cuboid ground truth at a scale hand annotation cannot reach — the annotation set was extracted from professional Kujiale.com interior designs, not labelled by hand. The ECCV 2020 acceptance and the authors' reported gains from mixing the data with real images are the real endorsements. Compare it with real-capture sets like ScanNet or Matterport3D when sensor realism matters: those give you noise and scanner artifacts, Structured3D gives you clean labels and no capture artifacts. Budget for domain adaptation either way.
Verdict generatedVerified 6d ago · liveness 56/100 · cite: rightaichoice.com/tools/structured3d
- Computer vision researchers training room layout estimation models
- ML teams pretraining 3D structure detectors on dense geometric labels
- Groups needing line, plane and junction labels at hand-labelling scale
- Synthetic-to-real transfer studies combining renders with real benchmarks
- Projects needing real captured imagery, sensor noise or scanner artifacts
- Commercial work before a legal review of the Structured3D Terms of Use
- Teams wanting an actively updated dataset with recent releases
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip Structured3D if your evaluation set is real scanned or photographed interior data and you need sensor noise, occlusion and capture artifacts represented in training rather than a clean synthetic structure prior.
The data is licensed under the Structured3D Terms of Use while only the code is MIT — plan legal review time if your work is commercial.
Structured3D costs nothing and has no tiers, so there is no pricing ladder to climb. The real comparison is effort: free but gated behind an agreement form and a self-written dataloader, versus paid commercial scene datasets or in-house capture and labelling, where dense line and plane annotation is the expensive line item. For a lab already paying annotators, the design-derived labels here are the cost saver.
In short
Structured3D — Free synthetic indoor-scene dataset with photo-realistic renders and dense 3D structure annotations for layout research. Best for Computer vision researchers training room layout estimation models, ML teams pretraining 3D structure detectors on dense geometric labels, Groups needing line, plane and junction labels at hand-labelling scale. Free to use.
What is Structured3D?
Structured3D — Free synthetic indoor-scene dataset with photo-realistic renders and dense 3D structure annotations for layout research. Best for Computer vision researchers training room layout estimation models, ML teams pretraining 3D structure detectors on dense geometric labels, Groups needing line, plane and junction labels at hand-labelling scale. Free to use.
Structured3D is a research dataset, not a product. It ships 3.5K house designs made by professional interior designers at Kujiale.com, with ground-truth 3D structure annotations extracted automatically from those design files, plus photo-realistic 2D renders generated from an industry-standard rendering engine. Because the annotations come out of the design data rather than human labelling, the set scales to millions of individual structure instances — junctions, line segments, planes, cuboids, room layouts and per-instance 3D bounding boxes. Both panoramic and perspective image views are included, and a small MIT-licensed code package lets you view the annotations. The paper, by Zheng, Zhang, Li, Tang, Gao and Zhou, was accepted to ECCV 2020. In it the authors trained room-layout-estimation networks on Structured3D combined with real images and reported improved performance on benchmark datasets. You get access by filling out an agreement form accepting the Structured3D Terms of Use; the download link follows once the form is received. It suits computer-vision researchers and ML engineers working on room layout estimation, holistic scene understanding, 3D structure detection and synthetic-to-real transfer. The trade-off is domain gap: synthetic renders do not behave like scanner captures, so treat it as a training or pretraining resource and validate on real imagery before shipping.
Key features
- 3.5K house designs created by professional interior designers
- Automatic 3D structure extraction from interior design files
- Photo-realistic renders from an industry-standard rendering engine
- Panoramic image views for 360-degree scene training
- Perspective image views for standard camera-model training
- Junction annotations for structural corner detection
- Line segment annotations at scale
- Plane annotations for surface and wall reconstruction
- Cuboid annotations for volumetric structure
- Room layout annotations for layout estimation benchmarks
- Per-instance 3D bounding boxes
- MIT-licensed code for viewing structure annotations
- Synthetic-to-real experiments combining the data with real images
- Semantic labels for 3D bounding boxes via the related Ctrl-Room dataset
- Additional panoramic views via the related PNVS dataset
Behind the verdict
Structured3D is a dataset release, and reading it as anything else will mislead you. There is no dashboard, no account system beyond the agreement form, and no managed pipeline — the site's News entries stop at July 2020, when the ECCV 2020 acceptance was announced, so plan on a fixed snapshot rather than an actively versioned resource. What you actually get is unusually well-matched to a specific class of problem. The annotations — junctions, line segments, planes, cuboids, room layouts and per-instance 3D bounding boxes — fall out of the underlying interior design files automatically. That is the whole argument: line segments number in the millions, and getting comparable labels out of scanned data means expensive, noisy manual work. Both panoramic and perspective views ship, which covers 360-degree training and standard camera-model training from one source. The MIT-licensed visualization code is genuinely basic, so expect to write your own dataloaders rather than inherit a training pipeline. Where it fits: pretraining structure detectors, benchmarking holistic scene understanding methods, multi-task learning on paired 2D image and 3D structure, novel view synthesis on the panoramic set (the related PNVS release adds more panoramic views for exactly that), and instance-level 3D box detection with semantic labels available from the related Ctrl-Room release. The ECCV 2020 workshop challenges gave the dataset an external benchmarking venue, which matters if you want comparable numbers rather than your own splits. Where it does not fit: anything needing real capture characteristics — sensor noise, motion blur, scanner artifacts, real-world clutter. If your evaluation set is real photography or scan data, the synthetic-to-real gap is your first engineering problem, not an afterthought. Commercial use needs a look at the Structured3D Terms of Use, since the data is licensed under those terms while only the code is MIT; the house designs originate with Kujiale.com, which the authors thank for the design database and rendering engine. The practical sequence: submit the agreement form, wait for access, then validate a small model end-to-end on a real benchmark before you commit a training budget. If the geometry supervision is what you need, few free alternatives offer this density.
Limitations
- Structured3D is a static research dataset, not an interactive tool.
- The site's News entries run from August 2019 to July 2020, ending with the ECCV 2020 acceptance, so do not expect ongoing feature or version support.
- Access is gated: you submit the agreement form accepting the Structured3D Terms of Use and receive download access after it is received.
- The bundled code is basic annotation visualization under the MIT license, not a training pipeline — the data itself is under the Structured3D Terms of Use, so commercial use needs a review.
- Images are synthetic, so models trained on them alone usually need domain adaptation before real-world use.
- There is no live service to build around: no dashboard, API or managed workflow.
as of 2026-10-20
Researching Structured3D? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Is Structured3D still active?
Structured3D is live · free · last verified 24 Oct 2026 · viability monitor.
Viability score
How well maintained and how widely used is Structured3D? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →What do users say about Structured3D?
What people actually say about Structured3D — is it worth it?
Our community sentiment research written Jul 30, 2026 is more than 90 days old, so it is being re-reviewed against current data and is not shown.
How we verified this
Every listed fact about Structured3D— pricing, features, links, status — is re-read from the vendor's own pages by our verification pipeline and each change is logged with a date; the last pass ran on 24 Oct 2026. Pages mix model-assisted synthesis with human review; see how we verify.
Verification history
We have re-verified Structured3D 9 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — changed features, use cases, description, pricing fit, who should skip it, hidden costs, pricing tiers, switching out, limitations, summary, our verdict, switching in, who it suits, setup time
- — changed description, who it suits, who should skip it, summary, our verdict, features, pricing tiers
- — changed use cases, who it suits, description, our verdict, features, hidden costs, limitations, switching in, switching out, who should skip it, pricing tiers, pricing fit, setup time, summary
- — changed limitations
- — changed features, pricing tiers, summary
- — changed SSL certificate, domain expiry
Showing the 6 most recent of 9 verification passes.
Free to cite with attribution — this page re-verifies continuously.
How much does Structured3D cost in 2026?
Structured3D: Free to use. Pricing last verified 24 Oct 2026.
Plans compared
Every published Structured3D tier, its price as the vendor publishes it, who it fits, and what it adds over the tier below. The 12-month cost projection follows.
| Plan | Price | Billing | Who it's for |
|---|---|---|---|
| Free (agreement form required) | Free | — | Academic and industry researchers who need dense 3D structure ground truth and can accept a research Terms of Use plus a form-based access step.Adds: Free entry point: 3.5K designs, panoramic and perspective renders, full annotation set, and MIT-licensed visualization code. |
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Where the pricing makes sense
The company stage and team size where Structured3D's pricing actually pencils out — and where peers do it cheaper.
Structured3D costs nothing and has no tiers, so there is no pricing ladder to climb. The real comparison is effort: free but gated behind an agreement form and a self-written dataloader, versus paid commercial scene datasets or in-house capture and labelling, where dense line and plane annotation is the expensive line item. For a lab already paying annotators, the design-derived labels here are the cost saver.
Has Structured3D's pricing changed?
No plan price for Structured3D has moved since we began logging in Aug 2026; the pricing page is re-read on every verification pass.
Who should pick Structured3D?
Pick Structured3D if you are Computer vision researchers training room layout estimation models, ML teams pretraining 3D structure detectors on dense geometric labels, Groups needing line, plane and junction labels at hand-labelling scale, Synthetic-to-real transfer studies combining renders with real benchmarks.
Look elsewhere if you are Projects needing real captured imagery, sensor noise or scanner artifacts, Commercial work before a legal review of the Structured3D Terms of Use, Teams wanting an actively updated dataset with recent releases, Anyone unwilling to submit the agreement form and wait for access.
Skip it if Skip Structured3D if your evaluation set is real scanned or photographed interior data and you need sensor noise, occlusion and capture artifacts represented in training rather than a clean synthetic structure prior.
GeneratedReal-world workflow fit
Concrete scenarios for the personas Structured3D actually fits — and what changes day-one when you adopt it.
Submit the agreement form, download the panoramic and perspective images, and wire the junction, line and plane annotations into a layout network using the MIT visualization code as a reference for the annotation format.
Outcome: A layout model pretrained on dense synthetic geometry, then fine-tuned and validated on a real benchmark dataset to quantify the synthetic-to-real gap.
Use the per-instance 3D bounding boxes for instance-level detection pretraining, pulling semantic labels from the related Ctrl-Room release to give the boxes classes.
Outcome: A detector pretrained on millions of automatically extracted structure instances before fine-tuning on the team's real capture data.
Evaluate holistic scene structure methods against the junction, line, plane and cuboid labels and compare against the Holistic 3D Vision Challenges venue hosted at the ECCV 2020 workshop.
Outcome: Numbers that are comparable with published work rather than a private split, with the ECCV 2020 paper as the reference baseline.
Use cases
- Train deep networks for room layout estimation with dense structural ground truth.
- Benchmark holistic scene understanding models against junction, line, plane and cuboid labels.
- Build paired 2D image and 3D structure data for multi-task learning.
- Run novel view synthesis experiments on the panoramic image set, optionally with PNVS views.
- Develop synthetic-to-real transfer pipelines by mixing Structured3D with real images.
- Evaluate 3D bounding box detection with per-instance boxes, using Ctrl-Room for semantic labels.
- Compete or benchmark on the Holistic 3D Vision Challenges run at the ECCV 2020 workshop.
Setup time & first value
How long it actually takes to get something useful out of Structured3D — broken out by persona, not the marketing-page minute.
Allow an afternoon for the agreement form and access turnaround, then a day or two to write a dataloader against the annotation format — the MIT code shows you the schema but not a training loop. Expect first real benchmark numbers only after adding a domain-adaptation pass on real imagery.
Switching to or from Structured3D
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From ScanNet: swap noisy captured geometry for clean design-derived line and plane labels, then keep ScanNet as the real-world validation set.
- →From Matterport3D: pair Structured3D's dense structure annotations with panoramic renders for pretraining, and use Matterport3D for real sensor evaluation.
- →From in-house labelled captures: pretrain on the 3.5K designs to cut manual annotation volume, then fine-tune on your own labelled subset.
- ↗To ScanNet: move to real captured geometry when scanner artifacts and noise become the signal your model must learn.
- ↗To Matterport3D: switch for real-world panoramic captures at production fidelity.
- ↗To a commercial scene dataset: move when you need contractually clear commercial licensing and vendor support rather than a research Terms of Use.
Resources & guides
Tutorials & Learning
YouTube returned 6 videos for “Structured3D”, and we withheld 6: 6 could not be judged, because “Structured3D” is a single word that other videos use for other things. We are showing none, because we could not prove any of them are about Structured3D.
Official links
What are the best Structured3D alternatives?
The closest alternatives to Structured3D we list are Datagen, Markup and Fashion Mnist. Ranked by direct product-type match, not category overlap.
| Alternative | Pricing | Starts at | Last verified |
|---|---|---|---|
| Datagen | Contact Sales | Custom pricing | — |
| Markup | Freemium | Pricing not published | — |
| Fashion Mnist | Free | Free | — |
Alternatives to Structured3D
View all
Datagen
Synthetic data generation for computer vision that ships photorealistic, auto-annotated training sets on demand.
Markup
Markup turns PDFs and long text into structured NLP and ML datasets with AI-assisted annotation in the browser
Fashion Mnist
A free 70,000-image Zalando clothing dataset built as a drop-in replacement for MNIST.
Tools that pair well with Structured3D
Common stack mates teams adopt alongside Structured3D, with the specific reason each pairing earns its keep.
Datagen
Synthetic data generation for computer vision that ships photorealistic, auto-annotated training sets on demand.
Markup
Markup turns PDFs and long text into structured NLP and ML datasets with AI-assisted annotation in the browser
Fashion Mnist
A free 70,000-image Zalando clothing dataset built as a drop-in replacement for MNIST.
Structured3D vs its closest rival
Structured3D's closest head-to-head is with Praktika — If you're a language learner aiming to improve spoken fluency, Praktika's AI tutors and real-time corrections offer a convenient mobile solution. If you're a researcher in 3D computer vision, Structured3D's free synthetic dataset with rich annotations is invaluable. These tools serve completely different needs—choose based on your domain. Verdict generated
Head-to-head comparisons
Structured3d vs Praktika
If you're a language learner aiming to improve spoken fluency, Praktika's AI tutors and real-time corrections offer a convenient mobile solution. If you're a researcher in 3D computer vision, Structured3D's free synthetic dataset with rich annotations is invaluable. These tools serve completely different needs—choose based on your domain.
GeneratedStructured3d vs Surge Ai
Choose Structured3D if you're a researcher needing a free, high-quality synthetic dataset with rich 3D annotations for scene understanding. Choose Surge AI if you're building or aligning frontier AI models and require expert human feedback, RLHF, or rigorous red teaming—especially for tasks like document understanding where benchmarks like GDP.pdf are used by top labs.
GeneratedFrequently asked questions
Best-of guides
Topics
Used Structured3D? Help shape our editorial sentiment research.