Can Chen, Ph.D. is a senior R&D engineer working remotely on Kitware’s Computer Vision Team. She focuses on developing tools for object-based detection, feature detection, motion pattern learning, and anomaly detection.

Prior to working at Kitware, Can was an R&D engineer for DGene. She was the lead algorithm developer for a deep learning-based SmartRoto system. SmartRoto aimed to alleviate manual labor for movie post-production by at least 50% using high-quality annotation. Users could select one or more targets with bounding boxes in a video, then output the tracked matte for each frame. She developed a new foreground segmentation network and a background extraction network and proposed an instance segmentation network that improved the segment boundary accuracy.

During her time at DGene, Can also was the team leader for a deep learning-based face reenactment project. She generated a 3D face representation from a 2D image for both the source actor from a single image. Then, the source’s expressions were deformation transferred to enact the target’s 3D face.

Can was a graduate intern at Honeywell while she was working on her Ph.D. She was involved in the DARPA MediFor project, where she detected manipulations introduced by the artificial blurring of images. She developed a deep learning framework for detecting and localizing forgeries (splicing and retouching) in photos via camera response function analysis. She also proposed a novel CNN framework to distinguish images that had a naturally shallow depth of field from images that were manipulated (e.g. iPhone portrait mode with dual-lens). Can built a new dataset of 178 camera response functions from modern digital cameras to analyze camera response functions, identify different camera sources, and evaluate camera response function modeling methods.

Can received her Ph.D. in computer science from the University of Delaware. She received both her master’s and bachelor’s degrees in electrical engineering from Harbin Institute of Technology.