Jacob Lin

I am a second-year CS PhD student at the University of Oxford, supervised by Ronald Clark and funded by EPSRC. My research interests broadly lie in Computer Vision, specifically in World Modelling and Physics-Guided Vision Models.

I have previously worked as a research assistant at Imperial College London, where I also did my undergraduate degree. Prior to my studies in the UK, I was born and raised in Malmö, Sweden.

Email  /  CV  /  Twitter  /  Github

profile photo

News

[09/2025] Cloud4D accepted to NeurIPS 2025 (Spotlight)!
[07/2025] Cloud-Stereo accepted to BMVC 2025
[10/2024] Started my PhD at Oxford!

Research

My research focuses on teaching vision models to understand physics. Currently, I'm approaching this through world modelling on large-scale data. Previously, I approached it through physics-based simulations, with a focus on atmospheric clouds.

Cloud4D overview Cloud4D: Estimating Cloud Properties at a High Spatial and Temporal Resolution
Jacob Lin, Edward Gryspeerdt, Ronald Clark
NeurIPS 2025 (Spotlight)
project page / paper

We use large-eddy simulations to train a vision model to estimate cloud properties from ground-based cameras, at 25 m and 5 s resolution.


Cloud-Stereo: A Dataset and Benchmark for Reconstructing Atmospheric Clouds from Stereo Images
Jacob Lin, Edward Gryspeerdt, Ronald Clark
BMVC 2025
project page / paper

We release a stereo dataset co-located with LiDAR for benchmarking 3D reconstruction of atmospheric clouds.


Template taken from here.