Oscar Carlsson
Papers
1
Total Citations
8
H-Index
1
About
Oscar Carlsson is a researcher at the forefront of geometric deep learning and computer vision, with a particular focus on spherical representations for wide-angle imagery. His most notable contribution is the development of HEAL-SWIN, a vision transformer architecture designed to operate directly on the sphere, circumventing the distortion and projection losses that plague conventional neural networks when processing fisheye and omnidirectional images. This work, published in 2024 and already garnering 8 citations, addresses a critical bottleneck in robotics and autonomous driving, where high-resolution, wide-angle perception is essential. By leveraging the HEALPix pixelation scheme, Carlsson’s approach enables efficient and distortion-free learning on spherical data, offering a principled alternative to equirectangular projections. His research bridges the gap between theoretical geometry and practical deployment, making him a rising figure in the field. As spherical vision gains traction in autonomous systems, Carlsson’s work is poised to become a foundational reference for researchers seeking to process wide-angle imagery without sacrificing accuracy or computational efficiency.
Research Focus
Key Achievements
Top Papers
- 1HEAL-SWIN: A Vision Transformer on the Sphere8 citations · 2024