MatSynth is a large-scale, publicly available dataset of 4,069 high-quality, 4K, tileable PBR materials with permissive licenses. The dataset includes 683,592 augmented materials with rotations and crops, and 3,417,960 renderings of the augmented materials under various illumination conditions. Each material is represented by seven textures: base color, diffuse, normal, height, roughness, metallic, and specular. The dataset also includes metadata such as origin, license, category, tags, creation method, and physical size. The materials are sourced from various online libraries and have been carefully curated to ensure they are non-duplicate, tileable, and of high quality. The dataset is designed to support modern learning-based techniques for a variety of material-related tasks, including material acquisition, material generation, and synthetic data generation. The dataset is released through the project page at: https://www.gvecchio.com/matsynth. The MatSynth dataset was created to address the limitations of existing material datasets, which are often small, limited in diversity, and not publicly available. The dataset includes a wide range of materials, including architectural, interior, luxury, and marble materials. The dataset is compatible with both Diffuse/Specular and Base Color/Metallic workflows. The materials are processed to ensure they are tileable and have the same set of parameters. The dataset also includes a large number of renderings of the materials under various environmental illuminations, enabling future learning and research. The dataset was created by collecting materials from multiple online sources under CC0 and CC-BY licensing frameworks. The materials were carefully selected and processed to ensure they are of high quality and non-duplicate. The dataset includes a wide range of materials, including those from AmbientCG, CGBookCase, PolyHeaven, ShateTexture, and TextureCan. The dataset also includes materials from the artist Julio Sillet. The materials were augmented with rotations, crops, and different environmental illuminations to facilitate future learning. The dataset has been evaluated for its impact on material-related tasks, including material acquisition and generation. The results show that the additional data enables higher quality acquisition and improved diversity in the generation results. The dataset has been used to train and evaluate various methods for material acquisition and generation, including Deschaintre et al. [5], SurfaceNet [43], and MatFuse [45]. The results show that the additional data significantly improves the performance of these methods. The dataset is also useful for synthetic data generation, for example, for object acquisition, material-based selection, and numerous applications further from material understanding and authoring. The dataset is a significant step towards reducing the gap between publicly and privately available material datasets, facilitating future research. The materials in the dataset are tileable and in 4K resolution, enabling future research on ultra-high resolution for materials. This is of particular interest as mostMatSynth is a large-scale, publicly available dataset of 4,069 high-quality, 4K, tileable PBR materials with permissive licenses. The dataset includes 683,592 augmented materials with rotations and crops, and 3,417,960 renderings of the augmented materials under various illumination conditions. Each material is represented by seven textures: base color, diffuse, normal, height, roughness, metallic, and specular. The dataset also includes metadata such as origin, license, category, tags, creation method, and physical size. The materials are sourced from various online libraries and have been carefully curated to ensure they are non-duplicate, tileable, and of high quality. The dataset is designed to support modern learning-based techniques for a variety of material-related tasks, including material acquisition, material generation, and synthetic data generation. The dataset is released through the project page at: https://www.gvecchio.com/matsynth. The MatSynth dataset was created to address the limitations of existing material datasets, which are often small, limited in diversity, and not publicly available. The dataset includes a wide range of materials, including architectural, interior, luxury, and marble materials. The dataset is compatible with both Diffuse/Specular and Base Color/Metallic workflows. The materials are processed to ensure they are tileable and have the same set of parameters. The dataset also includes a large number of renderings of the materials under various environmental illuminations, enabling future learning and research. The dataset was created by collecting materials from multiple online sources under CC0 and CC-BY licensing frameworks. The materials were carefully selected and processed to ensure they are of high quality and non-duplicate. The dataset includes a wide range of materials, including those from AmbientCG, CGBookCase, PolyHeaven, ShateTexture, and TextureCan. The dataset also includes materials from the artist Julio Sillet. The materials were augmented with rotations, crops, and different environmental illuminations to facilitate future learning. The dataset has been evaluated for its impact on material-related tasks, including material acquisition and generation. The results show that the additional data enables higher quality acquisition and improved diversity in the generation results. The dataset has been used to train and evaluate various methods for material acquisition and generation, including Deschaintre et al. [5], SurfaceNet [43], and MatFuse [45]. The results show that the additional data significantly improves the performance of these methods. The dataset is also useful for synthetic data generation, for example, for object acquisition, material-based selection, and numerous applications further from material understanding and authoring. The dataset is a significant step towards reducing the gap between publicly and privately available material datasets, facilitating future research. The materials in the dataset are tileable and in 4K resolution, enabling future research on ultra-high resolution for materials. This is of particular interest as most