{"url":"/dataset/matsynth","name":"MatSynth","full_name":null,"description_markdown":"# MatSynth\r\n\r\nMatSynth is a Physically Based Rendering (PBR) materials dataset designed for modern AI applications.\r\nThis dataset consists of over 4,000 ultra-high resolution, offering unparalleled scale, diversity, and detail. \r\n\r\nMeticulously collected and curated, MatSynth is poised to drive innovation in material acquisition and generation applications, providing a rich resource for researchers, developers, and enthusiasts in computer graphics and related fields.\r\n\r\n## Dataset Description\r\n\r\nMatSynth is a new large-scale dataset comprising over 4,000 ultra-high resolution Physically Based Rendering (PBR) materials, \r\nall released under permissive licensing.\r\n\r\nAll materials in the dataset are represented by a common set of maps (*Basecolor*, *Diffuse*, *Normal*, *Height*, *Roughness*, *Metallic*, *Specular* and, when useful, *Opacity*), \r\nmodelling both the reflectance and mesostructure of the material.\r\n\r\nEach material in the dataset comes with rich metadata, including information on its origin, licensing details, category, tags, creation method, \r\nand, when available, descriptions and physical size. \r\nThis comprehensive metadata facilitates precise material selection and usage, catering to the specific needs of users.\r\n\r\n## Dataset Structure\r\n \r\nThe MatSynth dataset is divided into two splits: the test split, containing 89 materials, and the train split, consisting of 3,980 materials. \r\nTo enhance accessibility and ease of navigation, each split is further organized into separate folders for each distinct category present in the dataset (Blends, Ceramic, Concrete, Fabric, Ground, Leather, Marble, Metal, Misc, Plastic, Plaster, Stone, Terracotta, Wood). \r\n\r\n## Dataset Creation\r\n\r\nThe MatSynth dataset is designed to support modern, learning-based techniques for a variety of material-related tasks including, \r\nbut not limited to, material acquisition, material generation and synthetic data generation e.g. for retrieval or segmentation. \r\n\r\n## Source Data\r\n\r\nThe MatSynth dataset is the result of an extensively collection of data from multiple online sources operating under the CC0 and CC-BY licensing framework. \r\nThis collection strategy allows to capture a broad spectrum of materials, \r\nfrom commonly used ones to more niche or specialized variants while guaranteeing that the data can be used for a variety of usecases. \r\n\r\nMaterials under CC0 license were collected from [AmbientCG](https://ambientcg.com/), [CGBookCase](https://www.cgbookcase.com/), [PolyHeaven](https://polyhaven.com/), \r\n[ShateTexture](https://www.sharetextures.com/), and [TextureCan](https://www.texturecan.com/).\r\nThe dataset also includes limited set of materials from the artist [Julio Sillet](https://juliosillet.gumroad.com/), distributed under CC-BY license.\r\n\r\nWe collected over 6000 materials which we meticulously filter to keep only tileable, 4K materials. \r\nThis high resolution allows us to extract many different crops from each sample at different scale for augmentation. \r\nAdditionally, we discard blurry or low-quality materials (by visual inspection). \r\nThe resulting dataset consists of 3736 unique materials which we augment by blending semantically compatible materials (e.g.: snow over ground). \r\nIn total, our dataset contains 4069 unique 4K materials.\r\n\r\n## Annotations\r\n\r\nThe dataset is composed of material maps (Basecolor, Diffuse, Normal, Height, Roughness, Metallic, Specular and, when useful, opacity) \r\nand associated renderings under varying environmental illuminations, and multi-scale crops.\r\nWe adopt the OpenGL standard for the Normal map (Y-axis pointing upward). \r\nThe Height map is given in a 16-bit single channel format for higher precision.\r\n\r\nIn addition to these maps, the dataset includes other annotations providing context to each material: \r\nthe capture method (photogrammetry, procedural generation, or approximation); \r\nlist of descriptive tags; source name (website); source link; \r\nlicensing and a timestamps for eventual future versioning. \r\nFor a subset of materials, when the information is available, we also provide the author name (387), text description (572) and a physical size, \r\npresented as the length of the edge in centimeters (358). \r\n\r\n## Citation\r\n\r\n```\r\n@article{vecchio2024matsynth,\r\n  title={MatSynth: A Modern PBR Materials Dataset},\r\n  author={Vecchio, Giuseppe and Deschaintre, Valentin},\r\n  journal={arXiv preprint arXiv:2401.06056},\r\n  year={2024}\r\n}\r\n```","description_withheld":null,"homepage":"https://gvecchio.com/matsynth/","introduced_date":"2024-01-11","introduced_date_note":null,"introduced_by":{"paper":"/paper/matsynth-a-modern-pbr-materials-dataset","title":"MatSynth: A Modern PBR Materials Dataset","first_author":"Giuseppe Vecchio","url":null},"license":null,"modalities":[{"name":"Images","url":"/datasets/modality/images"},{"name":"Texts","url":"/datasets/modality/texts"},{"name":"3D","url":"/datasets/modality/3d"}],"tasks":[{"name":"Image Generation","url":"/task/image-generation","datasets_with_task":"/datasets/task/image-generation"},{"name":"SVBRDF Estimation","url":"/task/svbrdf-estimation","datasets_with_task":"/datasets/task/svbrdf-estimation"}],"languages":[{"name":"English","url":"/datasets/language/english"}],"variants":["MatSynth"],"data_loaders":[{"repo":"https://github.com/huggingface/datasets","url":"https://huggingface.co/datasets/gvecchio/MatSynth","frameworks":["tf","pytorch","jax"]}],"num_papers_in_archive":5,"source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28"},"benchmarks":[],"papers_with_a_benchmark_row":[],"syntology_totals":{"read_at":"2026-09-24T18:15:14+00:00","papers_with_samples":0,"samples_harvested":0,"samples_ran":0,"samples_unverified":0,"pointer_only_for_licence":0,"papers_with_no_sample_that_ran":0,"note":"the per-paper counts above, summed; not a rate"},"papers_note":"The archive never published its papers-using-dataset list; these are papers with a leaderboard row on this dataset's benchmarks."}