Dataset News
The latest Dataset coverage curated for game developers.
arXiv cs.GR details SsgCaps, a public dataset for benchmarking sound scene generation. The set uses human-engineered scenes tied to structured prompts and only public-domain audio, making it easier for teams …
SyntheticDoc brings 1,000,000 procedurally generated, high-resolution training images to document unwarping and illumination correction. The dataset pairs each sample with pixel-perfect UV, normal, albedo, and shading maps, aiming to replace …
A new motion-generation system aims to cover both humans and animals with one model, sidestepping the topology mismatch that usually forces separate pipelines. UniMo pairs a point-cloud representation with dynamic …
SeMoCo reframes motion tokens around meaning first, splitting each token into a semantic core and residual kinematics. For animation and ML pipelines, that could make text-to-motion generation easier to steer …
A robotics team has built a bathing-demo pipeline that captures motion, body shape, contact regions, and force in sync, then transfers that data to a soft robotic hand. The system …
SubtleTalk tackles a long-standing problem in audio-driven 3D talking heads: believable motion above the mouth. The system adds controllable eyebrow, blink, head, and affect cues on top of lip sync …
A new CG-World dataset packages 850,000 temporally aligned 1-5 second segments from industrial graphics pipelines into a world-state format. It records latent state, motion, camera, lighting, physics, and events, giving …
A 106-hour, 764-speaker dataset is the main practical contribution here, because it lets gesture models be tested with speaker-disjoint splits instead of accidentally learning one person’s style. The paper argues …
A 100,000-sample CAD dataset now ships with executable Python programs, validated feature histories, STEP output, point clouds, and renderings. For tools and technical teams, the interesting part is that each …
A new music-driven dance generation paper claims state-of-the-art results by moving the problem into a latent space and conditioning on music genre. The practical takeaway for animation and tools teams …
A new reconstruction method tackles a nasty edge case for graphics teams: scenes that change across long time spans instead of moving smoothly. ChronoGS uses a temporally modulated Gaussian representation …
A new 3D idle-motion dataset could fill a real gap for character animation teams: 50 subjects, about 6 hours of captured idle behavior, plus a baseline model and evaluation code. …
A new 109,495-panorama dataset could matter for teams working on 3D reconstruction, capture pipelines, or VR-style scene understanding. Holo360D pairs panoramas with registered point clouds, meshes, and aligned poses, and …
A new dataset for complex 4D markerless human motion capture has been introduced, addressing the limitations of traditional marker-based systems. This dataset captures intricate human interactions, making it invaluable for …
A new dataset, FIT, introduces over 1.13 million virtual try-on images, addressing the critical issue of garment fit in virtual fashion. This development is particularly relevant for graphics programmers and …
The introduction of the MAROON dataset marks a significant advancement in imaging techniques, combining high-resolution radio-frequency and optical depth imaging. This innovation is particularly relevant for graphics programmers and technical …
A new CAD benchmark puts editability, not just shape match, at the center of evaluation. HistCAD packages 170,236 executable parametric sequences with explicit constraints, STEP files, renders, and text labels. …
VibraVerse introduces a groundbreaking dataset that aligns geometry and acoustics for multimodal learning. This dataset enhances the understanding of physical properties and their acoustic responses, crucial for game developers focusing …