# 3D Scene Reconstruction

The work redesigns the object-centric 3D generator Trellis 2 with adaptive chunks that scale with camera distance, explicit 2D–3D conditioning that lifts image features onto observed surfaces while distinguishing free space from occluded regions, and supervision from about 4,000 synthesized outdoor scenes, generating a complete scene mesh from a single image and outperforming all baselines in geometric accuracy and perceptual quality on Tanks and Temples, ScanNet++, and 121 in-the-wild images.

The work redesigns the object-centric 3D generator Trellis 2 with adaptive chunks that scale with camera distance, explicit 2D–3D conditioning that lifts image features onto observed surfaces while distinguishing free space from occluded regions, and supervision from about 4,000 synthesized outdoor scenes, generating a complete scene mesh from a single image and outperforming all baselines in geometric accuracy and perceptual quality on Tanks and Temples, ScanNet++, and 121 in-the-wild images.

Mira-Scene is a compositional single-image 3D scene reconstruction framework that pairs a pixel-aligned Canonical Coordinate Map (CCM) with a scene-space Point Cloud Map (PCM) from monocular geometry estimation to form dense, bounded correspondences, recovers object transformations through robust geometric alignment, and jointly generates object geometry and CCMs with a multimodal diffusion transformer; across indoor, outdoor, synthetic, and in-the-wild scenes it improves layout accuracy over strong baselines, achieving relative gains of 39.8% in 3D-IoU and 16.5% in 2D-IoU over SAM3D on BlendSwap while using limited open-source training data.

Mira-Scene is a compositional single-image 3D scene reconstruction framework that pairs a pixel-aligned Canonical Coordinate Map (CCM) with a scene-space Point Cloud Map (PCM) from monocular geometry estimation to form dense, bounded correspondences, recovers object transformations through robust geometric alignment, and jointly generates object geometry and CCMs with a multimodal diffusion transformer; across indoor, outdoor, synthetic, and in-the-wild scenes it improves layout accuracy over strong baselines, achieving relative gains of 39.8% in 3D-IoU and 16.5% in 2D-IoU over SAM3D on BlendSwap while using limited open-source training data.