publications

2024

1. 

   FineControlNet: Fine-level Text Control for Image Generation with Spatially Aligned Text Control Injection

   Hongsuk Choi, Isaac Kasahara, Selim Engin, and 3 more authors

   2024

   Desc arXiv Website

   Text and pose embeddings can be composed using attention masks during the diffusion process to achieve instance-level control when generating images.

2023

1. 

   Real-time Simultaneous Multi-Object 3D Shape Reconstruction, 6DoF Pose Estimation and Dense Grasp Prediction

   Shubham Agrawal, Nikhil Chavan-Dafle, Isaac Kasahara, and 3 more authors

   2023

   Desc arXiv Website

   Simultaneous multi-object pose, shape, and grasp estimation can allow inference in real-time for robotic manipulation.

2. 

   RIC: Rotate-Inpaint-Complete for Generalizable Scene Reconstruction

   Isaac Kasahara, Shubham Agrawal, Selim Engin, and 3 more authors

   2023

   Desc arXiv Website

   Text-to-image models pre-trained on large amounts of image data can be used to perform generalizable 3D scene reconstruction from a single viewpoint.

2022

1. 

   Look Both Ways: Self-supervising Driver Gaze Estimation and Road Scene Saliency

   Isaac Kasahara, Simon Stent, and Hyun Soo Park

   In Computer Vision – ECCV 2022, 2022

   Desc PDF Website

   Utilizing saliency information from outside the vehicle can improve the attention estimation of the driver in a self-supervised fashion.