Thursday, September 19, 2024 12:30pm to 1:30pm
About this Event
170 S University St, West Lafayette, IN 47907-2072
https://ag.purdue.edu/events/department/fnr/2024/09/data-driven-seminar-series-dr-bedrich-benes.htmlDr. Bedrich Benes, a Professor of Computer Science and the Associate Department Head for Purdue University, will be conducting a seminar on Generative AI for Digital Twin Reconstruction.
When: September 19th from 12:30pm - 1:30pm
Hosted By: Institute of Digital Forestry in Whistler Hall (WSLR) room 116
Generative AI for Digital Twin Reconstruction
Tree reconstruction has undergone unprecedented progress in the past 40 years. It has evolved from matching simple cone-like static shapes to high- detailed physics-enabled models with photonic-level illumination simulating plant growth with complex intrinsic and extrinsic signaling at interactive framerates. These models are currently being complemented with generative AI. This talk will discuss several methods for plant reconstruction AI. In particular, we will show how trees can be localized and counted from satellite images in urban layouts using deep-learning-based inverse neurosymbolic procedural models that learn the planting rules. We will also show how tree geometry can be learned and encoded into a compact representation using generative AI and how a diffusion model approximated 100,000 tree geometries from Google Street maps using single photographs.
Bedrich Benes is a professor and associate head of Computer Science at Purdue University and a Purdue faculty scholar. He leads the Computational Vegetation group at CS. He is a Eurographics Fellow and IEEE and ACM senior member. Professor Benes is the editor-in-chief of Elsevier Graphical Models. He has worked as an international program committee member of various conferences, including Siggraph, Siggraph Asia, Eurographics, and Symposium of Geometric Modeling. He is one of the editors of In Silico Plants journal by Oxford Publishing. Dr. Benes works in generative methods for geometry synthesis, and his main focus is procedural and inverse procedural modeling, deep learning, and simulation of natural phenomena. He has published over 250 research papers in the field. Bedrich has been sponsored by the National Science Foundation, NASA, Adobe Research, Intel, Siemens, Samsung, the Department of Energy, and Ford Inc. His graduate students work in Adobe Research, Google, Microsoft, Nvidia, Samsung, Siemens, and Square Enix.
Since Fall 2020, Purdue Agriculture has hosted the Data-Driven Agriculture seminar series featuring experts in data science and digital agriculture from across Purdue's College of Agriculture and beyond.
To view all past seminars, check out the Data-Driven Agriculture playlist on Purdue Agriculture’s YouTube channel.