Modality Forcing for Scalable Spatial Generation
2026-06-11 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors introduce Modality Forcing, a straightforward method to train models that generate both images and their depth maps together using limited depth data. Their approach uses separate noise settings for image and depth parts, allowing flexible and joint creation of both outputs. They found that bigger models trained on more image data improve depth prediction accuracy, with their best model matching current leading techniques. This shows that training on image generation tasks can also help machines understand 3D space better.
Text-to-Image ModelsDepth PredictionModality ForcingSparse Depth DataDiT ModelConditional GenerationJoint Image-Depth GenerationNoise LevelsMonocular Depth EstimationSpatial Perception
Authors
Bardienus Pieter Duisterhof, Deva Ramanan, Jeffrey Ichnowski, Justin Johnson, Keunhong Park
Abstract
Text-to-image (T2I) models contain rich spatial priors. Synthesizing photorealistic, cluttered scenes requires an understanding of geometry, including perspective and relative scale. Prior works adapt T2I models to leverage this prior for depth prediction, but they require dense depth data and involve complex recipes. We propose Modality Forcing, a simple, scalable post-training recipe for joint image-depth generation using a single DiT trained on sparse depth data. Modality Forcing enables conditional and joint generation of image and depth in any permutation by assigning separate noise levels per modality. Per-modality decoders let us train on sparse, real-world depth and achieve strong, generalizable depth prediction. We further show that Modality Forcing inherits the scalability of T2I pre-training: by training a set of T2I models from scratch (370M to 3.3B parameters), we find that larger models trained on more image data produce more accurate depth. Our strongest model is competitive with state-of-the-art monocular depth estimators and reduces AbsRel by 57% relative to existing joint image-depth generative models. These results provide strong evidence that image generation is a scalable pre-training objective for spatial perception. https://modality-forcing.github.io/