Gunho Choi

Tomocube is a holotomography company which uses laser technique to measure three-dimensional refractive index (RI) tomogram of a microscopic sample such as biological cells and tissues. I apply deep learning techniques on RI tomogram such as bacteria cell classification, cell nucleus segmentation, unsupervised light coherence denoising, etc.

• Classification of Bacteria Cells
• Cell Nucleus Segmentation
• RI Tomogram Denoising

At Interpark, I collaborated with Severance hospital to detect and classify lesions that cause Epilepsy. For detecting lesions, I applied Anomaly Detection algorithms on 3D brain MRI volume on multiple protocols. For classification of Epilepsy, I took part in augmentation of EEG signals to handle data imbalance using InfoGAN.

• Anomaly Detection on MRI
• InfoGAN on EEG Singal

DeepBio is a company that applies DL on biological data. Classifying cancer state was one of the projects at the company and the classifier’s performance was degraded depending on the staining style of histopathological image which was different hospital by hospital. In order to overcome this issue, I worked on stain style transfer using DL technique called GAN.

• Stain Style Transfer of Histopathologic Images

Laftel is an animation streaming service in Korea. Simply, it is an animation version of Netflix. In Laftel, I worked on scene classification which was used for recommendation. Also, I worked on scene style clustering using DL based style extraction algorithm inspired by “A Neural Algorithm of Artistic Style”

• Scene Classification
• Animation Style Clustering

I served my military obligation in the United States Army as Korean Augmentation to the U.S. Army (KATUSA). I enlisted right after my freshman year. My job was to register biological information of the people who works in the garrison. “Gangnam Style” came out back then and my staff sergeant ordered me to dance to the music along the hallway.