SMPL-X
PulseAugur coverage of SMPL-X — every cluster mentioning SMPL-X across labs, papers, and developer communities, ranked by signal.
4 day(s) with sentiment data
-
New SRVQ Framework Enhances Sign Language Generation Accuracy
Researchers have developed a new framework called Semantic Residual Vector Quantization (SRVQ) for sign language generation. This method separates the process into two stages: first determining the semantic content of w…
-
New framework refines 3D human body shape models
Researchers have developed a new framework to refine 3D human body shape reconstructions from parametric models like SMPL. The framework combines three methods: Latent-VaRefine for correcting latent codes, Vertex-VaRefi…
-
New HAPACT benchmark targets physical impact localization in videos
Researchers have introduced HAPACT, a new benchmark and dataset designed to improve the localization of human-centric physical impacts in videos. This benchmark, derived from movie clips, focuses on identifying the temp…
-
New EgoExplorer dataset captures human loco-manipulation with wearable tech
Researchers have introduced EgoExplorer, a comprehensive 100-hour dataset capturing human loco-manipulation activities. This dataset utilizes a wearable system, including body trackers, a head-mounted GoPro, and special…
-
MegaAvatar framework generates controllable talking avatars with 3D motion guidance
Researchers have introduced MegaAvatar, a new framework for generating controllable talking avatars. This system builds upon the Wan2.2-TI2V-5B model and incorporates 3D guidance derived from SMPL-X data to control body…
-
EMODY Flow generates emotion-aware full-body motion from audio
Researchers have developed EMODY Flow, a novel framework for generating full-body motion that synchronizes with speech and emotional cues. This system addresses a limitation in existing models where emotion conditioning…
-
New AvaImg pipeline enhances 3D clothed human representation
Researchers have developed AvaImg, a novel multi-stage optimization pipeline designed to improve the representation of 3D clothed humans as standardized 2D UV texture and displacement maps. This method addresses the lon…
-
New benchmarks and models advance sign language translation accuracy · 3 sources tracked
Researchers are developing new benchmarks and models for sign language translation (SLT) to improve accuracy and capture nuances beyond traditional metrics. One paper proposes an LLM-based QA protocol to better assess s…
-
AniCrafter model enables realistic human animation in dynamic backgrounds
Researchers have developed AniCrafter, a novel diffusion-based model for human-centric animation. This model is designed to animate a character within dynamic, open-domain backgrounds while following specified human mot…
-
New GRAPE method enhances 3D portrait mesh estimation
Researchers have developed GRAPE, a new method for articulated portrait mesh estimation that addresses limitations in existing 3D Morphable Models. GRAPE introduces a Portrait Parametric Model with an explicit torso-to-…
-
FlexiAvatar framework generates 3D human avatars by optimizing visible body regions
Researchers have introduced FlexiAvatar, a novel framework for generating animatable 3D human avatars from monocular video. This method uniquely optimizes only the visible body regions, thereby preventing artifacts caus…
-
New framework improves sign language motion execution on humanoid robots
Researchers have developed a new framework to improve the execution of sign language generation (SLG) motions on humanoid robots. The system addresses issues like self-intersections and collisions in generated 3D body r…
-
New T3HG-Editor enables text-driven 3D human garment editing
Researchers have introduced T3HG-Editor, a novel system for text-driven 3D human garment editing. This approach addresses limitations in existing 3D Gaussian Editing methods, which often produce low-fidelity and inconsi…
-
Human4K dataset advances 3D human reconstruction with 4K multi-view mocap data · 2 sources tracked
Researchers have introduced Human4K, a new large-scale dataset designed to improve 3D human reconstruction. This dataset features over six million 4K images captured from an eight-view camera system, synchronized with p…
-
TryOnCrafter framework enables camera-controllable video virtual try-on · 3 sources tracked
Researchers have introduced TryOnCrafter, a novel framework for camera-controllable video virtual try-on. This system moves beyond existing methods by decoupling human subjects from their environments using a renderable…
-
VolHuMe dataset released with high-resolution 4D human scans
Researchers have introduced VolHuMe, a new dataset featuring high-resolution 4D human scans captured using 64 RGB and 32 depth cameras. This dataset includes detailed ground truth for 104 subjects, such as SMPL-X models…
-
PhysDrift framework generates physically executable co-speech motions for humanoid robots
Researchers have developed PhysDrift, a new framework designed to generate co-speech motions for humanoid robots that are both expressive and physically executable. The system addresses the "embodiment gap" by directly …
-
New system InHabit generates large-scale 3D human-scene interaction data
Researchers have developed InHabit, a novel system for generating large-scale, photorealistic 3D datasets of humans interacting with environments. This system leverages image foundation models to propose actions and ins…
-
DanceHMR recovers detailed human body and hand motion from video
Researchers have developed DanceHMR, a new framework for recovering detailed human body and hand movements from single-camera videos. This method aims to improve upon existing techniques by unifying body and hand motion…
-
Tamaththul3D creates high-fidelity 3D Saudi Sign Language avatars from video
Researchers have developed Tamaththul3D, a novel pipeline for generating high-fidelity 3D avatars of Saudi Sign Language (SSL). This system addresses a significant gap in resources for Arabic Sign Language (ArSL), which…