Search papers, labs, and topics across Lattice.
Sapiens2 is a family of high-resolution transformer models (0.4-5B parameters) designed for human-centric vision tasks, featuring native 1K and hierarchical 4K resolutions. The models are pretrained using a combined masked image reconstruction and self-distilled contrastive objective on a curated dataset of 1 billion high-quality human images. Sapiens2 achieves state-of-the-art results on pose estimation, body-part segmentation, normal estimation, and extends to new tasks like pointmap and albedo estimation, demonstrating improved generalization and high-fidelity outputs.
Human-centric vision gets a serious upgrade: Sapiens2 models smash previous benchmarks on pose, segmentation, and normal estimation by a significant margin, while also tackling new tasks like pointmap and albedo estimation.
We present Sapiens2, a model family of high-resolution transformers for human-centric vision focused on generalization, versatility, and high-fidelity outputs. Our model sizes range from 0.4 to 5 billion parameters, with native 1K resolution and hierarchical variants that support 4K. Sapiens2 substantially improves over its predecessor in both pretraining and post-training. First, to learn features that capture low-level details (for dense prediction) and high-level semantics (for zero-shot or few-label settings), we combine masked image reconstruction with self-distilled contrastive objectives. Our evaluations show that this unified pretraining objective is better suited for a wider range of downstream tasks. Second, along the data axis, we pretrain on a curated dataset of 1 billion high-quality human images and improve the quality and quantity of task annotations. Third, architecturally, we incorporate advances from frontier models that enable longer training schedules with improved stability. Our 4K models adopt windowed attention to reason over longer spatial context and are pretrained with 2K output resolution. Sapiens2 sets a new state-of-the-art and improves over the first generation on pose (+4 mAP), body-part segmentation (+24.3 mIoU), normal estimation (45.6% lower angular error) and extends to new tasks such as pointmap and albedo estimation. Code: https://github.com/facebookresearch/sapiens2
