Concerto: Joint 2D-3D Self-Supervised Learning Emerges Spatial Representations
This work addresses spatial cognition for scene understanding in AI, offering a novel approach that improves performance on benchmarks, though it is incremental in combining existing modalities.
The paper tackles the problem of learning spatial representations by introducing Concerto, a joint 2D-3D self-supervised learning method that simulates human concept learning, resulting in outperforming standalone SOTA 2D and 3D models by 14.2% and 4.8% in linear probing and achieving 80.7% mIoU on ScanNet with fine-tuning.
Humans learn abstract concepts through multisensory synergy, and once formed, such representations can often be recalled from a single modality. Inspired by this principle, we introduce Concerto, a minimalist simulation of human concept learning for spatial cognition, combining 3D intra-modal self-distillation with 2D-3D cross-modal joint embedding. Despite its simplicity, Concerto learns more coherent and informative spatial features, as demonstrated by zero-shot visualizations. It outperforms both standalone SOTA 2D and 3D self-supervised models by 14.2% and 4.8%, respectively, as well as their feature concatenation, in linear probing for 3D scene perception. With full fine-tuning, Concerto sets new SOTA results across multiple scene understanding benchmarks (e.g., 80.7% mIoU on ScanNet). We further present a variant of Concerto tailored for video-lifted point cloud spatial understanding, and a translator that linearly projects Concerto representations into CLIP's language space, enabling open-world perception. These results highlight that Concerto emerges spatial representations with superior fine-grained geometric and semantic consistency.