Sujin Jang

Title: Sujin Jang: Inventor in Data Labeling and Object Estimation

Introduction

Sujin Jang is an inventor based in Suwon, South Korea. Her patent work includes methods for preparing data to train machine-learning systems and for training models that estimate objects.

Latest Patents

Jang’s listed patents address data labeling and object-estimation model training.

- Method and Apparatus with Data Labeling. This invention describes obtaining localization information about an object, extracting a landmark point from a map, and using that point to generate a ground-truth image. The image is then refined to create training data.

- Method and Apparatus with Object Estimation Model Training. This invention describes training a neural-network model using a cross-correlation loss. The loss is generated from feature vectors produced by an interim model and a trained model, each receiving input data about a target object.

Career Highlights

Jang is associated with Samsung Electronics Ltd. Her listed portfolio includes 2 patents focused on data preparation and neural-network training for object estimation.

Collaborations

The provided information lists Dae Ung Jo and Yonggonjong Park as Jang’s coworkers. Their names appear alongside her patent-related work.

Conclusion

Sujin Jang’s listed inventions focus on generating refined training data and improving neural-network training for object estimation, reflecting her patent work associated with Samsung Electronics Ltd.