01
Start with an image.
A single view of the world. Our focus is monocular depth: estimating a scene’s spatial structure from one image.
A NEW PERSPECTIVE ON COMPUTER VISION
One image.
A deeper understanding.
Teaching machines to see beyond pixels.
We’re building monocular depth estimation guided by semantics — for creators, storytellers, and robots.
FROM IMAGE TO UNDERSTANDING+
SEMANTIC CONTEXT
SPATIAL STRUCTURE
THE WORLD HAS DEPTH.
EVERY IMAGE HOLDS A CLUE.Concept visualization
Not a model prediction
EVERY IMAGE HOLDS A CLUE.Concept visualization
Not a model prediction
MONOCULAR INPUT. SEMANTIC GUIDANCE. SPATIAL INTELLIGENCE.SCROLL TO EXPLORE
01 / OUR APPROACHMEANING MEETS GEOMETRY
Depth begins with
understanding.
A pixel tells you about color. Context helps you understand what it belongs to. We want to bring that understanding into the way machines estimate depth.
02 ✳
Give every pixel context.
ViT-Up upsamples Vision Transformer features at continuous image coordinates. We use these dense features as semantic guidance.
Meet ViT-Up03
Build a deeper picture.
Our goal is more coherent depth, with structure that respects objects and their boundaries — ready to inform creative and physical tasks.
Research in motion. We’re developing this approach and exploring where it can make a difference.
Read the foundational research02 / POSSIBILITIESFROM PIXELS TO PURPOSE
A new dimension
of possibility.
Better scene understanding can change how we create, imagine, and move. These are the applications we’re working toward.
REIMAGINE THE LIGHTFOR IMAGE MAKERS
Light, in a different way.
Explore depth-aware portrait relighting. Shape new light and shadow around the contours of a face.
IMAGE EDITING & RELIGHTING +
BRING EFFECTS INTO THE SCENEFOR STORYTELLERS
Make imagination fit.
Bring CGI into filmed reality. Explore depth-aware compositing, occlusion, and the integration of digital objects and effects.
FILM & VISUAL EFFECTS +
UNDERSTAND THE WORLDFOR EMBODIED INTELLIGENCE
See a way forward.
Investigate depth for robot perception — understanding nearby objects and the space around them from a camera image.
ROBOTICS & PERCEPTION +Not just seeing the world.
Making sense of it.
From a single image to a richer understanding of space.
That’s the future we’re building toward.
03 / THE FOUNDERSCURIOUS BY NATURE
Built on research.
Driven by possibility.
Two founders exploring the intersection of visual understanding, geometry, and the things it makes possible.
CO-FOUNDER
Igor Mihajlovic
MSc, Shanghai Jiao Tong University
CO-FOUNDER
Krispin Wandel
MSc, ETH Zurich
PhD candidate, Shanghai Jiao Tong University