{ "@context": { "@language": "en", "@vocab": "https://schema.org/", "citeAs": "cr:citeAs", "column": "cr:column", "conformsTo": "dct:conformsTo", "cr": "http://mlcommons.org/croissant/", "rai": "http://mlcommons.org/croissant/RAI/", "prov": "http://www.w3.org/ns/prov#", "data": { "@id": "cr:data", "@type": "@json" }, "dataType": { "@id": "cr:dataType", "@type": "@vocab" }, "dct": "http://purl.org/dc/terms/", "examples": { "@id": "cr:examples", "@type": "@json" }, "extract": "cr:extract", "field": "cr:field", "fileProperty": "cr:fileProperty", "fileObject": "cr:fileObject", "fileSet": "cr:fileSet", "format": "cr:format", "includes": "cr:includes", "isLiveDataset": "cr:isLiveDataset", "jsonPath": "cr:jsonPath", "key": "cr:key", "md5": "cr:md5", "parentField": "cr:parentField", "path": "cr:path", "recordSet": "cr:recordSet", "references": "cr:references", "regex": "cr:regex", "repeated": "cr:repeated", "replace": "cr:replace", "sc": "https://schema.org/", "separator": "cr:separator", "source": "cr:source", "subField": "cr:subField", "transform": "cr:transform" }, "@type": "sc:Dataset", "name": "IGF-Bench", "description": "IGF-Bench (Indoor Geometric Fidelity Benchmark) is the first benchmark for evaluating structural-level geometric fidelity of conditionally generated indoor scene images. Built atop 3D-FRONT, it provides 3,600 calibrated views from 300 rooms at three furniture-complexity levels (L0 empty, L1 basic, L2 full) with four camera viewpoints each. Six state-of-the-art Canny-conditioned diffusion models produce 21,600 generated images. Each view is accompanied by ground-truth metric depth (EXR), Canny edge maps, semantic segmentation masks, 3D wireframes, and camera parameters. Monocular depth estimates (DepthPro, with Depth Anything V2 and ZoeDepth ablations) enable four novel geometric fidelity metrics—planarity error (L_plane), orthogonality error (L_ortho), edge straightness (L_edge), and vanishing-point consistency (L_vp)—that capture structural degradation invisible to perceptual scores such as FID and LPIPS.", "conformsTo": "http://mlcommons.org/croissant/1.0", "version": "1.0.0", "datePublished": "2026-04-09", "dateCreated": "2026-03-01", "license": "https://creativecommons.org/licenses/by-nc-sa/4.0/", "url": "https://huggingface.co/datasets/igfbench-neurips2026/IGF-Bench", "sameAs": "https://huggingface.co/datasets/igfbench-neurips2026/IGF-Bench", "keywords": [ "geometric fidelity", "indoor scenes", "image generation", "ControlNet", "depth estimation", "3D-FRONT", "diffusion models", "benchmark", "monocular depth estimation", "vanishing points", "structural evaluation" ], "citeAs": "@inproceedings{igfbench2026, title={IGF-Bench: Evaluating Geometric Fidelity of Conditional Image Generation Beyond Perceptual Metrics}, author={Anonymous}, booktitle={NeurIPS 2026 Evaluations and Datasets Track}, year={2026}}", "creator": { "@type": "Organization", "name": "Anonymous (under double-blind review)" }, "publisher": { "@type": "Organization", "name": "Anonymous" }, "isLiveDataset": false, "isAccessibleForFree": true, "inLanguage": "en", "rai:dataCollection": "Ground-truth scenes are sourced from the 3D-FRONT dataset (Fu et al., 2021), a large-scale repository of professionally designed indoor layouts. 300 rooms were sampled across 7 subtypes (Bedroom 106, MasterBedroom 113, SecondBedroom 16, KidsRoom 12, ElderlyRoom 1, LivingRoom 18, LivingDiningRoom 34) and stratified into three composition groups by dominant type and footprint (Group A 177 rooms / Group B 94 / Group C 29). Selection criteria: complete ceiling-wall-floor enclosure, approximate floor area in 5--50 m^2, at least one window or door opening. Each room was rendered at three furniture-complexity levels (L0 empty, L1 basic, L2 fully furnished) from four calibrated camera viewpoints (placed at bounding-box corners, eye-height 1.43 m, FOV 90 degrees) using Blender 5.0 Cycles with physically based materials, yielding 3,600 ground-truth views at 1024x1024 resolution. Generated images were produced via Canny-conditioned ControlNet for six diffusion models (Stable Diffusion 1.5, SDXL, Stable Diffusion 3.5, Flux.1 Dev, Hunyuan-DiT, Kolors) under uniform settings: seed=42, cn_scale=1.0, Canny thresholds low=100/high=200, empty negative prompt. Depth estimation was performed with Apple DepthPro (primary), with ablation subsets using Depth Anything V2 and ZoeDepth. No crowd-sourcing or human annotation was involved; all labels are derived from the 3D scene geometry.", "rai:dataCollectionType": "Synthetically rendered from existing 3D assets; no new data collection from human participants.", "rai:personalSensitiveInformation": "None. The dataset contains no personal data, no human faces, no identifiable information, and no biometric data. All content is synthetically generated from 3D architectural scene models depicting indoor spaces with furniture.", "rai:dataBiases": "The 3D-FRONT source dataset consists of professionally designed Chinese residential interior layouts. Results may not generalize to: (1) non-Chinese architectural styles or culturally distinct interiors, (2) informal or low-income housing, (3) non-residential spaces (offices, hospitals, retail), (4) outdoor or mixed indoor-outdoor environments. All six generation models are Canny-conditioned; other conditioning types (depth, segmentation, sketch) are not evaluated. A single random seed (42) is used, limiting stochastic diversity analysis.", "rai:dataLimitations": "Stable Diffusion 1.5 generates at 512x512 resolution while the other five models generate at 1024x1024; cross-resolution comparison is noted as a limitation. Single seed (42) for all models. Monocular depth estimation is inherently scale-ambiguous and may introduce systematic errors. Synthetic ground truth may not capture all real-world material and lighting complexity. Only Canny edge conditioning is evaluated. Semantic masks are derived from render passes and may have minor aliasing artifacts at object boundaries.", "rai:dataUseCases": "Primary: benchmarking geometric fidelity of conditional image generation models for indoor scenes. Secondary: evaluating monocular depth estimation robustness across rendered vs. generated image domains; analysing ControlNet conditioning mechanism differences (channel concatenation vs. residual injection); architectural visualization quality assessment; 3D-aware image generation research; indoor scene understanding. Discouraged: do not use for generating photorealistic deceptive architectural renderings without explicit disclosure (the paper Section 6 notes that IGF-Bench could in principle facilitate such misuse, as with all generative-image benchmarks); do not use as the sole quality signal for production-grade 3D reconstruction without combining with held-out real-world data; do not use to evaluate generation outside the indoor-scene domain (the benchmark is restricted to enclosed indoor rooms).", "rai:dataSocialImpact": "This benchmark advances evaluation methodology for generative models in architecture and interior design. Potential positive impacts include improved quality control for AI-generated architectural imagery and more reliable synthetic data for downstream 3D reconstruction. Potential negative impacts are limited because the dataset contains only synthetic indoor scenes with no personal or sensitive content. Users should be aware that high-fidelity generated images could be misused for deceptive architectural renderings; the geometric metrics introduced here can serve as detection tools for such misuse.", "rai:annotationsPerItem": "Each of the 3,600 ground-truth views directory contains 9 files: (1) RGB render (rgb_textured.png, 1024x1024 8-bit), (2) metric depth map (depth.exr, OpenEXR 32-bit float, in metres; canonical source), (3) 8-bit depth visualisation (depth.png, QA only), (4) NumPy float32 cache of depth.exr (depth_gt.npy, ~4 MB; for fast loading without an EXR decoder during LoRA training), (5) Canny edge map (canny.png, thresholds 100/200), (6) semantic segmentation mask with class IDs (semantic_id.png, 0=background, 1=floor, 2=ceiling, 3=wall, 4=opening, 5=structural element, 6=furniture), (7) colour-mapped semantic visualisation (semantic_mask.png, QA only), (8) 3D wireframe overlay (wireframe_3d.png, QA only), (9) camera parameters (camera.json: position, target, FOV=90, bounding box). Each of the 21,600 generated images has a corresponding DepthPro depth estimate (NPY) under depth_results/gen//depthpro/. Ablation subsets include Depth Anything V2 and ZoeDepth estimates.", "rai:dataReleaseMaintenancePlan": "Maintained for at least 5 years on HuggingFace (igfbench-neurips2026/IGF-Bench) and on the anonymous code mirror at anonymous.4open.science/r/IGF-Bench-Code (camera-ready: GitHub). Bug reports and questions are received via the HuggingFace dataset card discussion tab and via GitHub Issues on the code repository. Versioning follows semantic versioning conventions: minor metadata or documentation fixes ship as 1.x patch releases, schema-breaking changes (e.g. new room types, additional generation models, additional metrics) bump the major version, and older versions remain accessible on HuggingFace as immutable revisions. No deprecation is planned within the 5-year maintenance window.", "rai:hasSyntheticData": true, "prov:wasGeneratedBy": { "@type": "prov:Activity", "name": "IGF-Bench synthetic rendering and generation pipeline", "description": "Ground-truth views were rendered from the 3D-FRONT scene corpus (Fu et al., 2021) at 1024x1024 using Blender 5.0 Cycles with physically based materials (4 calibrated viewpoints per room, FOV 90 degrees, eye-height 1.43 m). Canny edges were extracted from each GT render with thresholds 100/200 and used as ControlNet conditioning for six diffusion-based generation pipelines (Stable Diffusion 1.5, SDXL, Stable Diffusion 3.5, Flux.1 Dev, Hunyuan-DiT, Kolors) under uniform settings (seed 42, ControlNet strength 1.0, empty negative prompt). Monocular depth estimates were produced by Apple DepthPro (primary), with Depth Anything V2 and ZoeDepth as ablation alternatives. The pipeline is fully automated with deterministic seeds; no crowd workers or human annotators were involved. See DATASHEET.md Section 3 (Collection Process) for the complete provenance trace and the released code repository for runnable pipeline scripts." }, "distribution": [ { "@type": "cr:FileObject", "@id": "igf-bench-repository", "name": "igf-bench-repository", "description": "IGF-Bench dataset repository hosted on HuggingFace Datasets", "contentUrl": "https://huggingface.co/datasets/igfbench-neurips2026/IGF-Bench", "encodingFormat": "git+https", "sha256": "main" }, { "@type": "cr:FileSet", "@id": "gt-rgb-files", "name": "gt-rgb-files", "description": "Ground truth RGB renders (3,600 PNG images at 1024x1024)", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "image/png", "includes": "renders_textured/*/L*_*/view_*/rgb_textured.png" }, { "@type": "cr:FileSet", "@id": "gt-depth-files", "name": "gt-depth-files", "description": "Ground truth metric depth maps (3,600 EXR files, 32-bit float, metres)", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "image/x-exr", "includes": "renders_textured/*/L*_*/view_*/depth.exr" }, { "@type": "cr:FileSet", "@id": "gt-depth-png-files", "name": "gt-depth-png-files", "description": "Ground truth depth visualisations (3,600 PNG files)", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "image/png", "includes": "renders_textured/*/L*_*/view_*/depth.png" }, { "@type": "cr:FileSet", "@id": "gt-canny-files", "name": "gt-canny-files", "description": "Canny edge maps used as ControlNet conditioning input (3,600 PNG files)", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "image/png", "includes": "renders_textured/*/L*_*/view_*/canny.png" }, { "@type": "cr:FileSet", "@id": "gt-semantic-id-files", "name": "gt-semantic-id-files", "description": "Semantic class label maps (3,600 PNG files; pixel values: 0=background, 1=floor, 2=ceiling, 3=wall, 4=opening, 5=structural element, 6=furniture). Primary planar classes 1/2/3 are used for L_plane and L_ortho computation.", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "image/png", "includes": "renders_textured/*/L*_*/view_*/semantic_id.png" }, { "@type": "cr:FileSet", "@id": "gt-semantic-mask-files", "name": "gt-semantic-mask-files", "description": "Semantic colour mask visualisations (3,600 PNG files)", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "image/png", "includes": "renders_textured/*/L*_*/view_*/semantic_mask.png" }, { "@type": "cr:FileSet", "@id": "gt-wireframe-files", "name": "gt-wireframe-files", "description": "3D wireframe overlay renders (3,600 PNG files)", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "image/png", "includes": "renders_textured/*/L*_*/view_*/wireframe_3d.png" }, { "@type": "cr:FileSet", "@id": "gt-depth-npy-cache-files", "name": "gt-depth-npy-cache-files", "description": "Pre-computed NumPy float32 cache of depth.exr (3,600 NPY files, ~4 MB each, ~14 GB total). Provided for fast loading without an OpenEXR decoder, e.g. during LoRA fine-tuning. Numerically identical to depth.exr; the EXR file is the canonical source.", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "application/octet-stream", "includes": "renders_textured/*/L*_*/view_*/depth_gt.npy" }, { "@type": "cr:FileSet", "@id": "gt-camera-files", "name": "gt-camera-files", "description": "Camera parameters for each view (3,600 JSON files; FOV=90 degrees, position, target, bounding box)", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "application/json", "includes": "renders_textured/*/L*_*/view_*/camera.json" }, { "@type": "cr:FileSet", "@id": "gen-sd15-files", "name": "gen-sd15-files", "description": "Stable Diffusion 1.5 Canny-conditioned generated images (3,600 PNG, 512x512)", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "image/png", "includes": "generated/sd15_canny/*/L*_*/view_*/generated.png" }, { "@type": "cr:FileSet", "@id": "gen-sdxl-files", "name": "gen-sdxl-files", "description": "SDXL Canny-conditioned generated images (3,600 PNG, 1024x1024)", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "image/png", "includes": "generated/sdxl_canny/*/L*_*/view_*/generated.png" }, { "@type": "cr:FileSet", "@id": "gen-sd35-files", "name": "gen-sd35-files", "description": "Stable Diffusion 3.5 Canny-conditioned generated images (3,600 PNG, 1024x1024)", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "image/png", "includes": "generated/sd35_canny/*/L*_*/view_*/generated.png" }, { "@type": "cr:FileSet", "@id": "gen-flux1-files", "name": "gen-flux1-files", "description": "Flux.1 Dev Canny-conditioned generated images (3,600 PNG, 1024x1024)", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "image/png", "includes": "generated/flux1_canny/*/L*_*/view_*/generated.png" }, { "@type": "cr:FileSet", "@id": "gen-hunyuan-files", "name": "gen-hunyuan-files", "description": "Hunyuan-DiT Canny-conditioned generated images (3,600 PNG, 1024x1024)", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "image/png", "includes": "generated/hunyuan_canny/*/L*_*/view_*/generated.png" }, { "@type": "cr:FileSet", "@id": "gen-kolors-files", "name": "gen-kolors-files", "description": "Kolors Canny-conditioned generated images (3,600 PNG, 1024x1024)", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "image/png", "includes": "generated/kolors_canny/*/L*_*/view_*/generated.png" }, { "@type": "cr:FileSet", "@id": "gen-sd15-with-neg-files", "name": "gen-sd15-with-neg-files", "description": "Stable Diffusion 1.5 generated images with the legacy negative prompt (App C.3 ablation; 3,552 paired views = 296 rooms x 12 views).", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "image/png", "includes": "generated/sd15_canny_with_neg/sd15_canny/*/L*_*/view_*/generated.png" }, { "@type": "cr:FileSet", "@id": "gen-sdxl-with-neg-files", "name": "gen-sdxl-with-neg-files", "description": "SDXL generated images with the legacy negative prompt (App C.3 ablation; 3,552 paired views).", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "image/png", "includes": "generated/sdxl_canny_with_neg/sdxl_canny/*/L*_*/view_*/generated.png" }, { "@type": "cr:FileSet", "@id": "gen-sd15-upsampled-files", "name": "gen-sd15-upsampled-files", "description": "Stable Diffusion 1.5 outputs bicubic-upsampled to 1024x1024 (App C.4 resolution-confound ablation; 3,552 paired views).", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "image/png", "includes": "generated/sd15_upsampled/sd15_canny/*/L*_*/view_*/generated.png" }, { "@type": "cr:FileSet", "@id": "gen-flux1-seed123-files", "name": "gen-flux1-seed123-files", "description": "Flux.1 Dev generations with seed=123 on the 200-view subset (App C.6 seed sensitivity / App C.7 quality gating).", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "image/png", "includes": "generated/ablation/flux1_seed123/*/L*_*/view_*/generated.png" }, { "@type": "cr:FileSet", "@id": "gen-flux1-seed456-files", "name": "gen-flux1-seed456-files", "description": "Flux.1 Dev generations with seed=456 on the 200-view subset (App C.6 / C.7).", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "image/png", "includes": "generated/ablation/flux1_seed456/*/L*_*/view_*/generated.png" }, { "@type": "cr:FileSet", "@id": "gen-flux1-seed789-files", "name": "gen-flux1-seed789-files", "description": "Flux.1 Dev generations with seed=789 on the 200-view subset (App C.7 quality gating only).", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "image/png", "includes": "generated/ablation/flux1_seed789/*/L*_*/view_*/generated.png" }, { "@type": "cr:FileSet", "@id": "gen-flux1-seed1024-files", "name": "gen-flux1-seed1024-files", "description": "Flux.1 Dev generations with seed=1024 on the 200-view subset (App C.7 quality gating only).", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "image/png", "includes": "generated/ablation/flux1_seed1024/*/L*_*/view_*/generated.png" }, { "@type": "cr:FileSet", "@id": "gen-sdxl-w050-files", "name": "gen-sdxl-w050-files", "description": "SDXL generations with controlnet_conditioning_scale=0.50 on the 200-view subset (App C.7 quality gating).", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "image/png", "includes": "generated/ablation/sdxl_w050/*/L*_*/view_*/generated.png" }, { "@type": "cr:FileSet", "@id": "gen-sdxl-w075-files", "name": "gen-sdxl-w075-files", "description": "SDXL generations with controlnet_conditioning_scale=0.75 on the 200-view subset (App C.7 quality gating).", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "image/png", "includes": "generated/ablation/sdxl_w075/*/L*_*/view_*/generated.png" }, { "@type": "cr:FileSet", "@id": "gen-sdxl-w125-files", "name": "gen-sdxl-w125-files", "description": "SDXL generations with controlnet_conditioning_scale=1.25 on the 200-view subset (App C.7 quality gating).", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "image/png", "includes": "generated/ablation/sdxl_w125/*/L*_*/view_*/generated.png" }, { "@type": "cr:FileSet", "@id": "gen-sdxl-w150-files", "name": "gen-sdxl-w150-files", "description": "SDXL generations with controlnet_conditioning_scale=1.50 on the 200-view subset (App C.7 quality gating).", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "image/png", "includes": "generated/ablation/sdxl_w150/*/L*_*/view_*/generated.png" }, { "@type": "cr:FileSet", "@id": "depth-gt-depthpro-files", "name": "depth-gt-depthpro-files", "description": "DepthPro depth estimates on ground-truth renders (3,600 NPY files)", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "application/x-npy", "includes": "depth_results/gt/depthpro/*/L*_*/view_*.npy" }, { "@type": "cr:FileSet", "@id": "depth-gt-dav2-files", "name": "depth-gt-dav2-files", "description": "Depth Anything V2 depth estimates on ground-truth renders (ablation subset, NPY files)", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "application/x-npy", "includes": "depth_results/gt/dav2/*/L*_*/view_*.npy" }, { "@type": "cr:FileSet", "@id": "depth-gt-zoedepth-files", "name": "depth-gt-zoedepth-files", "description": "ZoeDepth depth estimates on ground-truth renders (ablation subset, NPY files)", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "application/x-npy", "includes": "depth_results/gt/zoedepth/*/L*_*/view_*.npy" }, { "@type": "cr:FileSet", "@id": "depth-gen-depthpro-files", "name": "depth-gen-depthpro-files", "description": "DepthPro depth estimates on generated images (6 models x 3,600 = 21,600 NPY files)", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "application/x-npy", "includes": "depth_results/gen/*/depthpro/*/L*_*/view_*.npy" }, { "@type": "cr:FileSet", "@id": "depth-gen-dav2-files", "name": "depth-gen-dav2-files", "description": "Depth Anything V2 depth estimates on generated images (ablation subset, NPY files)", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "application/x-npy", "includes": "depth_results/gen/*/dav2/*/L*_*/view_*.npy" }, { "@type": "cr:FileSet", "@id": "depth-gen-zoedepth-files", "name": "depth-gen-zoedepth-files", "description": "ZoeDepth depth estimates on generated images (ablation subset, NPY files)", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "application/x-npy", "includes": "depth_results/gen/*/zoedepth/*/L*_*/view_*.npy" }, { "@type": "cr:FileSet", "@id": "depth-gen-sd15-neg-depthpro-files", "name": "depth-gen-sd15-neg-depthpro-files", "description": "DepthPro depth estimates on the SD 1.5 + legacy-negative-prompt ablation generations (App C.3; 3,552 NPY files).", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "application/x-npy", "includes": "depth_results/gen/sd15_neg/depthpro/*/L*_*/view_*.npy" }, { "@type": "cr:FileSet", "@id": "depth-gen-sdxl-neg-depthpro-files", "name": "depth-gen-sdxl-neg-depthpro-files", "description": "DepthPro depth estimates on the SDXL + legacy-negative-prompt ablation generations (App C.3; 3,552 NPY files).", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "application/x-npy", "includes": "depth_results/gen/sdxl_neg/depthpro/*/L*_*/view_*.npy" }, { "@type": "cr:FileSet", "@id": "depth-gen-sd15-upsampled-depthpro-files", "name": "depth-gen-sd15-upsampled-depthpro-files", "description": "DepthPro depth estimates on the SD 1.5 bicubic-upsampled-to-1024 ablation generations (App C.4; 3,552 NPY files).", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "application/x-npy", "includes": "depth_results/gen/sd15_upsampled/depthpro/*/L*_*/view_*.npy" }, { "@type": "cr:FileSet", "@id": "depth-gen-flux1-seed123-depthpro-files", "name": "depth-gen-flux1-seed123-depthpro-files", "description": "DepthPro depth estimates on Flux.1 seed=123 ablation generations (App C.6 / C.7; 200 NPY files).", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "application/x-npy", "includes": "depth_results/gen/flux1_seed123/depthpro/*/L*_*/view_*.npy" }, { "@type": "cr:FileSet", "@id": "depth-gen-flux1-seed456-depthpro-files", "name": "depth-gen-flux1-seed456-depthpro-files", "description": "DepthPro depth estimates on Flux.1 seed=456 ablation generations (App C.6 / C.7; 200 NPY files).", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "application/x-npy", "includes": "depth_results/gen/flux1_seed456/depthpro/*/L*_*/view_*.npy" }, { "@type": "cr:FileSet", "@id": "depth-gen-flux1-seed789-depthpro-files", "name": "depth-gen-flux1-seed789-depthpro-files", "description": "DepthPro depth estimates on Flux.1 seed=789 ablation generations (App C.7 only; 200 NPY files).", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "application/x-npy", "includes": "depth_results/gen/flux1_seed789/depthpro/*/L*_*/view_*.npy" }, { "@type": "cr:FileSet", "@id": "depth-gen-flux1-seed1024-depthpro-files", "name": "depth-gen-flux1-seed1024-depthpro-files", "description": "DepthPro depth estimates on Flux.1 seed=1024 ablation generations (App C.7 only; 200 NPY files).", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "application/x-npy", "includes": "depth_results/gen/flux1_seed1024/depthpro/*/L*_*/view_*.npy" }, { "@type": "cr:FileSet", "@id": "depth-gen-sdxl-w050-depthpro-files", "name": "depth-gen-sdxl-w050-depthpro-files", "description": "DepthPro depth estimates on SDXL cn_scale=0.50 ablation generations (App C.7; 200 NPY files).", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "application/x-npy", "includes": "depth_results/gen/sdxl_w050/depthpro/*/L*_*/view_*.npy" }, { "@type": "cr:FileSet", "@id": "depth-gen-sdxl-w075-depthpro-files", "name": "depth-gen-sdxl-w075-depthpro-files", "description": "DepthPro depth estimates on SDXL cn_scale=0.75 ablation generations (App C.7; 200 NPY files).", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "application/x-npy", "includes": "depth_results/gen/sdxl_w075/depthpro/*/L*_*/view_*.npy" }, { "@type": "cr:FileSet", "@id": "depth-gen-sdxl-w125-depthpro-files", "name": "depth-gen-sdxl-w125-depthpro-files", "description": "DepthPro depth estimates on SDXL cn_scale=1.25 ablation generations (App C.7; 200 NPY files).", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "application/x-npy", "includes": "depth_results/gen/sdxl_w125/depthpro/*/L*_*/view_*.npy" }, { "@type": "cr:FileSet", "@id": "depth-gen-sdxl-w150-depthpro-files", "name": "depth-gen-sdxl-w150-depthpro-files", "description": "DepthPro depth estimates on SDXL cn_scale=1.50 ablation generations (App C.7; 200 NPY files).", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "application/x-npy", "includes": "depth_results/gen/sdxl_w150/depthpro/*/L*_*/view_*.npy" }, { "@type": "cr:FileObject", "@id": "evaluation-igf-results", "name": "evaluation-igf-results", "description": "Per-view IGF evaluation results (L_plane, L_ortho, L_edge, L_vp for all models and views)", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "application/json", "contentUrl": "evaluation/igf_results.json" }, { "@type": "cr:FileObject", "@id": "evaluation-igf-summary", "name": "evaluation-igf-summary", "description": "Aggregated IGF evaluation summary statistics per model and complexity level", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "application/json", "contentUrl": "evaluation/igf_summary.json" }, { "@type": "cr:FileObject", "@id": "evaluation-fid-lpips", "name": "evaluation-fid-lpips", "description": "FID and LPIPS perceptual evaluation results per model", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "application/json", "contentUrl": "evaluation/fid_lpips.json" }, { "@type": "cr:FileObject", "@id": "evaluation-error-decomposition", "name": "evaluation-error-decomposition", "description": "Triangulated error decomposition (MDE error vs. generation error) per model", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "application/json", "contentUrl": "evaluation/error_decomposition.json" }, { "@type": "cr:FileObject", "@id": "evaluation-significance", "name": "evaluation-significance", "description": "Statistical significance tests (ANOVA, pairwise comparisons) for metric differences", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "application/json", "contentUrl": "evaluation/wilcoxon.json" }, { "@type": "cr:FileObject", "@id": "evaluation-mde-ablation", "name": "evaluation-mde-ablation", "description": "MDE ablation study results comparing DepthPro, Depth Anything V2, and ZoeDepth", "containedIn": { "@id": "igf-bench-repository" }, "encodingFormat": "application/json", "contentUrl": "evaluation/mde_ablation_results.json" } ], "recordSet": [ { "@type": "cr:RecordSet", "@id": "calibrated-views-rgb", "name": "calibrated-views-rgb", "description": "3,600 calibrated GT views indexed by (room_id, level, view) with the corresponding rgb_textured.png payload. Each row is one of the 300 rooms x 3 complexity levels x 4 viewpoints = 3,600 per-view files.", "field": [ { "@type": "cr:Field", "@id": "calibrated-views-rgb/room_id", "name": "room_id", "description": "3D-FRONT room identifier (UUID + RoomType-SceneID), extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-rgb-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/([^/]+)/L[012]_[a-z]+/view_[0-3]/rgb_textured\\.png$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-rgb/level", "name": "level", "description": "Furniture complexity level: L0_empty (bare room), L1_basic (key furniture only), or L2_full (fully furnished). Extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-rgb-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/[^/]+/(L[012]_[a-z]+)/view_[0-3]/rgb_textured\\.png$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-rgb/view", "name": "view", "description": "Camera viewpoint index (view_0 through view_3, placed at room bounding-box corners at 1.43 m eye height with FOV 90 degrees). Extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-rgb-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/[^/]+/L[012]_[a-z]+/(view_[0-3])/rgb_textured\\.png$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-rgb/rgb", "name": "rgb", "description": "Ground-truth RGB render (1024x1024 PNG); the primary visual input fed to Canny extraction and to all six generation pipelines as paired GT.", "dataType": "sc:ImageObject", "source": { "fileSet": { "@id": "gt-rgb-files" } } } ] }, { "@type": "cr:RecordSet", "@id": "calibrated-views-depth-exr", "name": "calibrated-views-depth-exr", "description": "3,600 calibrated GT views indexed by (room_id, level, view) with the corresponding depth.exr payload. Each row is one of the 300 rooms x 3 complexity levels x 4 viewpoints = 3,600 per-view files.", "field": [ { "@type": "cr:Field", "@id": "calibrated-views-depth-exr/room_id", "name": "room_id", "description": "3D-FRONT room identifier (UUID + RoomType-SceneID), extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-depth-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/([^/]+)/L[012]_[a-z]+/view_[0-3]/depth\\.exr$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-depth-exr/level", "name": "level", "description": "Furniture complexity level: L0_empty (bare room), L1_basic (key furniture only), or L2_full (fully furnished). Extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-depth-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/[^/]+/(L[012]_[a-z]+)/view_[0-3]/depth\\.exr$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-depth-exr/view", "name": "view", "description": "Camera viewpoint index (view_0 through view_3, placed at room bounding-box corners at 1.43 m eye height with FOV 90 degrees). Extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-depth-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/[^/]+/L[012]_[a-z]+/(view_[0-3])/depth\\.exr$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-depth-exr/depth_exr", "name": "depth_exr", "description": "Ground-truth metric depth map (32-bit OpenEXR float, in metres). Canonical depth source used for L_plane / L_ortho / L_edge baselines and for LoRA training supervision.", "dataType": "sc:URL", "source": { "fileSet": { "@id": "gt-depth-files" } } } ] }, { "@type": "cr:RecordSet", "@id": "calibrated-views-depth-png", "name": "calibrated-views-depth-png", "description": "3,600 calibrated GT views indexed by (room_id, level, view) with the corresponding depth.png payload. Each row is one of the 300 rooms x 3 complexity levels x 4 viewpoints = 3,600 per-view files.", "field": [ { "@type": "cr:Field", "@id": "calibrated-views-depth-png/room_id", "name": "room_id", "description": "3D-FRONT room identifier (UUID + RoomType-SceneID), extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-depth-png-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/([^/]+)/L[012]_[a-z]+/view_[0-3]/depth\\.png$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-depth-png/level", "name": "level", "description": "Furniture complexity level: L0_empty (bare room), L1_basic (key furniture only), or L2_full (fully furnished). Extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-depth-png-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/[^/]+/(L[012]_[a-z]+)/view_[0-3]/depth\\.png$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-depth-png/view", "name": "view", "description": "Camera viewpoint index (view_0 through view_3, placed at room bounding-box corners at 1.43 m eye height with FOV 90 degrees). Extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-depth-png-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/[^/]+/L[012]_[a-z]+/(view_[0-3])/depth\\.png$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-depth-png/depth_png", "name": "depth_png", "description": "8-bit normalised depth visualisation (PNG, QA only). Not used for evaluation; provided for quick visual inspection.", "dataType": "sc:ImageObject", "source": { "fileSet": { "@id": "gt-depth-png-files" } } } ] }, { "@type": "cr:RecordSet", "@id": "calibrated-views-depth-npy", "name": "calibrated-views-depth-npy", "description": "3,600 calibrated GT views indexed by (room_id, level, view) with the corresponding depth_gt.npy payload. Each row is one of the 300 rooms x 3 complexity levels x 4 viewpoints = 3,600 per-view files.", "field": [ { "@type": "cr:Field", "@id": "calibrated-views-depth-npy/room_id", "name": "room_id", "description": "3D-FRONT room identifier (UUID + RoomType-SceneID), extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-depth-npy-cache-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/([^/]+)/L[012]_[a-z]+/view_[0-3]/depth_gt\\.npy$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-depth-npy/level", "name": "level", "description": "Furniture complexity level: L0_empty (bare room), L1_basic (key furniture only), or L2_full (fully furnished). Extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-depth-npy-cache-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/[^/]+/(L[012]_[a-z]+)/view_[0-3]/depth_gt\\.npy$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-depth-npy/view", "name": "view", "description": "Camera viewpoint index (view_0 through view_3, placed at room bounding-box corners at 1.43 m eye height with FOV 90 degrees). Extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-depth-npy-cache-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/[^/]+/L[012]_[a-z]+/(view_[0-3])/depth_gt\\.npy$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-depth-npy/depth_gt_npy", "name": "depth_gt_npy", "description": "NumPy float32 cache of depth.exr (~4 MB per view). Numerically identical to depth.exr; provided for fast loading in environments without an OpenEXR decoder (e.g. during LoRA fine-tuning).", "dataType": "sc:URL", "source": { "fileSet": { "@id": "gt-depth-npy-cache-files" } } } ] }, { "@type": "cr:RecordSet", "@id": "calibrated-views-canny", "name": "calibrated-views-canny", "description": "3,600 calibrated GT views indexed by (room_id, level, view) with the corresponding canny.png payload. Each row is one of the 300 rooms x 3 complexity levels x 4 viewpoints = 3,600 per-view files.", "field": [ { "@type": "cr:Field", "@id": "calibrated-views-canny/room_id", "name": "room_id", "description": "3D-FRONT room identifier (UUID + RoomType-SceneID), extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-canny-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/([^/]+)/L[012]_[a-z]+/view_[0-3]/canny\\.png$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-canny/level", "name": "level", "description": "Furniture complexity level: L0_empty (bare room), L1_basic (key furniture only), or L2_full (fully furnished). Extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-canny-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/[^/]+/(L[012]_[a-z]+)/view_[0-3]/canny\\.png$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-canny/view", "name": "view", "description": "Camera viewpoint index (view_0 through view_3, placed at room bounding-box corners at 1.43 m eye height with FOV 90 degrees). Extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-canny-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/[^/]+/L[012]_[a-z]+/(view_[0-3])/canny\\.png$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-canny/canny", "name": "canny", "description": "Canny edge map (PNG; thresholds low=100/high=200). Used as the uniform ControlNet conditioning input for all six generation pipelines.", "dataType": "sc:ImageObject", "source": { "fileSet": { "@id": "gt-canny-files" } } } ] }, { "@type": "cr:RecordSet", "@id": "calibrated-views-semantic-id", "name": "calibrated-views-semantic-id", "description": "3,600 calibrated GT views indexed by (room_id, level, view) with the corresponding semantic_id.png payload. Each row is one of the 300 rooms x 3 complexity levels x 4 viewpoints = 3,600 per-view files.", "field": [ { "@type": "cr:Field", "@id": "calibrated-views-semantic-id/room_id", "name": "room_id", "description": "3D-FRONT room identifier (UUID + RoomType-SceneID), extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-semantic-id-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/([^/]+)/L[012]_[a-z]+/view_[0-3]/semantic_id\\.png$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-semantic-id/level", "name": "level", "description": "Furniture complexity level: L0_empty (bare room), L1_basic (key furniture only), or L2_full (fully furnished). Extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-semantic-id-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/[^/]+/(L[012]_[a-z]+)/view_[0-3]/semantic_id\\.png$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-semantic-id/view", "name": "view", "description": "Camera viewpoint index (view_0 through view_3, placed at room bounding-box corners at 1.43 m eye height with FOV 90 degrees). Extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-semantic-id-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/[^/]+/L[012]_[a-z]+/(view_[0-3])/semantic_id\\.png$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-semantic-id/semantic_id", "name": "semantic_id", "description": "Semantic class label map (PNG; pixel value = class ID with 0=background, 1=floor, 2=ceiling, 3=wall, 4=opening, 5=structural element, 6=furniture). Primary planar classes 1/2/3 are used for L_plane and L_ortho computation; furniture (class 6) is excluded.", "dataType": "sc:ImageObject", "source": { "fileSet": { "@id": "gt-semantic-id-files" } } } ] }, { "@type": "cr:RecordSet", "@id": "calibrated-views-semantic-rgb", "name": "calibrated-views-semantic-rgb", "description": "3,600 calibrated GT views indexed by (room_id, level, view) with the corresponding semantic_mask.png payload. Each row is one of the 300 rooms x 3 complexity levels x 4 viewpoints = 3,600 per-view files.", "field": [ { "@type": "cr:Field", "@id": "calibrated-views-semantic-rgb/room_id", "name": "room_id", "description": "3D-FRONT room identifier (UUID + RoomType-SceneID), extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-semantic-mask-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/([^/]+)/L[012]_[a-z]+/view_[0-3]/semantic_mask\\.png$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-semantic-rgb/level", "name": "level", "description": "Furniture complexity level: L0_empty (bare room), L1_basic (key furniture only), or L2_full (fully furnished). Extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-semantic-mask-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/[^/]+/(L[012]_[a-z]+)/view_[0-3]/semantic_mask\\.png$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-semantic-rgb/view", "name": "view", "description": "Camera viewpoint index (view_0 through view_3, placed at room bounding-box corners at 1.43 m eye height with FOV 90 degrees). Extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-semantic-mask-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/[^/]+/L[012]_[a-z]+/(view_[0-3])/semantic_mask\\.png$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-semantic-rgb/semantic_mask", "name": "semantic_mask", "description": "Colour-mapped visualisation of semantic_id.png (RGB PNG, QA only). Not used by any evaluation script.", "dataType": "sc:ImageObject", "source": { "fileSet": { "@id": "gt-semantic-mask-files" } } } ] }, { "@type": "cr:RecordSet", "@id": "calibrated-views-wireframe", "name": "calibrated-views-wireframe", "description": "3,600 calibrated GT views indexed by (room_id, level, view) with the corresponding wireframe_3d.png payload. Each row is one of the 300 rooms x 3 complexity levels x 4 viewpoints = 3,600 per-view files.", "field": [ { "@type": "cr:Field", "@id": "calibrated-views-wireframe/room_id", "name": "room_id", "description": "3D-FRONT room identifier (UUID + RoomType-SceneID), extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-wireframe-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/([^/]+)/L[012]_[a-z]+/view_[0-3]/wireframe_3d\\.png$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-wireframe/level", "name": "level", "description": "Furniture complexity level: L0_empty (bare room), L1_basic (key furniture only), or L2_full (fully furnished). Extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-wireframe-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/[^/]+/(L[012]_[a-z]+)/view_[0-3]/wireframe_3d\\.png$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-wireframe/view", "name": "view", "description": "Camera viewpoint index (view_0 through view_3, placed at room bounding-box corners at 1.43 m eye height with FOV 90 degrees). Extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-wireframe-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/[^/]+/L[012]_[a-z]+/(view_[0-3])/wireframe_3d\\.png$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-wireframe/wireframe_3d", "name": "wireframe_3d", "description": "3D wireframe overlay render (PNG, QA only). Provided for visual debugging of camera calibration; not used by IGF metrics.", "dataType": "sc:ImageObject", "source": { "fileSet": { "@id": "gt-wireframe-files" } } } ] }, { "@type": "cr:RecordSet", "@id": "calibrated-views-camera", "name": "calibrated-views-camera", "description": "3,600 calibrated GT views indexed by (room_id, level, view) with the corresponding camera.json payload. Each row is one of the 300 rooms x 3 complexity levels x 4 viewpoints = 3,600 per-view files.", "field": [ { "@type": "cr:Field", "@id": "calibrated-views-camera/room_id", "name": "room_id", "description": "3D-FRONT room identifier (UUID + RoomType-SceneID), extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-camera-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/([^/]+)/L[012]_[a-z]+/view_[0-3]/camera\\.json$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-camera/level", "name": "level", "description": "Furniture complexity level: L0_empty (bare room), L1_basic (key furniture only), or L2_full (fully furnished). Extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-camera-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/[^/]+/(L[012]_[a-z]+)/view_[0-3]/camera\\.json$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-camera/view", "name": "view", "description": "Camera viewpoint index (view_0 through view_3, placed at room bounding-box corners at 1.43 m eye height with FOV 90 degrees). Extracted from the file path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "gt-camera-files" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "^(?:.*/)?renders_textured/[^/]+/L[012]_[a-z]+/(view_[0-3])/camera\\.json$" } } }, { "@type": "cr:Field", "@id": "calibrated-views-camera/camera", "name": "camera", "description": "Per-view camera intrinsics and extrinsics (JSON: position, target, FOV=90, bounding box). Required to project depth into 3D point clouds for L_plane / L_ortho computation.", "dataType": "sc:URL", "source": { "fileSet": { "@id": "gt-camera-files" } } } ] }, { "@type": "cr:RecordSet", "@id": "generated-images", "name": "generated-images", "description": "21,600 generated images produced by six Canny-conditioned diffusion models (6 models x 3,600 views).", "field": [ { "@type": "cr:Field", "@id": "generated-images/room_id", "name": "room_id", "description": "3D-FRONT room identifier, matching calibrated-views/room_id", "dataType": "sc:Text" }, { "@type": "cr:Field", "@id": "generated-images/level", "name": "level", "description": "Furniture complexity level (L0_empty, L1_basic, L2_full)", "dataType": "sc:Text" }, { "@type": "cr:Field", "@id": "generated-images/view", "name": "view", "description": "Camera viewpoint index (view_0 through view_3)", "dataType": "sc:Text" }, { "@type": "cr:Field", "@id": "generated-images/model", "name": "model", "description": "Generation model identifier: sd15 (Stable Diffusion 1.5, 512x512), sdxl (SDXL, 1024x1024), sd35 (Stable Diffusion 3.5, 1024x1024), flux1 (Flux.1 Dev, 1024x1024), hunyuan (Hunyuan-DiT, 1024x1024), kolors (Kolors, 1024x1024)", "dataType": "sc:Text" }, { "@type": "cr:Field", "@id": "generated-images/image", "name": "image", "description": "Canny-conditioned generated image (PNG; 512x512 for sd15, 1024x1024 for all others; seed=42, cn_scale=1.0, empty negative prompt)", "dataType": "sc:ImageObject", "source": { "fileSet": { "@id": "gen-sd15-files" } } } ] }, { "@type": "cr:RecordSet", "@id": "depth-estimates", "name": "depth-estimates", "description": "Monocular depth estimates for ground-truth renders and generated images. Primary estimator: DepthPro (25,200 files). Ablation estimators: Depth Anything V2 and ZoeDepth (subsets).", "field": [ { "@type": "cr:Field", "@id": "depth-estimates/room_id", "name": "room_id", "description": "3D-FRONT room identifier", "dataType": "sc:Text" }, { "@type": "cr:Field", "@id": "depth-estimates/level", "name": "level", "description": "Furniture complexity level", "dataType": "sc:Text" }, { "@type": "cr:Field", "@id": "depth-estimates/view", "name": "view", "description": "Camera viewpoint index", "dataType": "sc:Text" }, { "@type": "cr:Field", "@id": "depth-estimates/image_source", "name": "image_source", "description": "Whether depth was estimated from a ground-truth render (gt) or a generated image (gen)", "dataType": "sc:Text" }, { "@type": "cr:Field", "@id": "depth-estimates/gen_model", "name": "gen_model", "description": "Generation model name (sd15, sdxl, sd35, flux1, hunyuan, kolors); only populated when image_source=gen", "dataType": "sc:Text" }, { "@type": "cr:Field", "@id": "depth-estimates/mde_method", "name": "mde_method", "description": "Monocular depth estimation method: depthpro (primary), dav2 (Depth Anything V2, ablation), zoedepth (ZoeDepth, ablation)", "dataType": "sc:Text" }, { "@type": "cr:Field", "@id": "depth-estimates/depth_npy", "name": "depth_npy", "description": "Estimated depth map stored as a NumPy array (.npy, float32)", "dataType": "sc:URL", "source": { "fileSet": { "@id": "depth-gen-depthpro-files" } } } ] } ] }