The challenge in CT imaging is that the data collected is often insufficient for high-resolution images, leading to noise in the measurements.

In my postdoc, I shifted towards inverse imaging problems, where the observed data is not a direct image but a transform, such as in CT scans.

When reconstructing 3D objects from line integrals, it's important to remember that the measurements are line integrals of the object you want to reconstruct.

Lidar measurements provide 3D models of trees, allowing for more detailed analysis compared to traditional planar images.

By integrating a 3D representation into models, we can better align the model's capabilities with the tasks we want it to perform, enhancing user affordances. Spatial intelligence is crucial for creating technology that understands and interacts with the 3D world, enabling a wide range of applications.

In using neural networks for problems like computer tomography, we are exploring how to combine handcrafted models with neural networks, particularly in how we feed them prior information.

Denoising is integrated into the image reconstruction process, addressing both the lack of data and the noise present in measurements.

In collaborations with hospitals, we focus on developing algorithms that maximize the quality of high-resolution images from limited data, particularly in medical imaging.

We're using drones to analyze the built environment, focusing on processing lines rather than points, which is traditionally more challenging in computer vision.