Anuj Sharma

Title: Anuj Sharma: Inventor in Sensor-Data Feature Extraction

Introduction

Anuj Sharma is an inventor based in Cambridge, Great Britain. His patent work focuses on machine-learning methods for extracting useful features from sensor data. He is associated with Five AI Limited and has 2 patents.

Latest Patents

Both of Sharma’s listed patents concern extracting features from sensor data and training machine-learning encoders.

One patent describes training an encoder alongside a perception component using both unannotated and annotated sensor data. Training alternates between two phases: first, the encoder is updated using a self-supervised objective and unannotated data; second, task-specific parameters are updated using annotated data and a task-specific objective. The approach aims to learn sensor-data features while training the perception component for a desired task.

The other patent presents a self-supervised method for training an encoder to associate two representations of a common set of sensor data. At least one representation is generated by using objects detected in an associated image to transform the sensor data. The learning objective encourages the system to associate the paired representations based on the features extracted from them.

Career Highlights

Sharma’s listed inventions address complementary aspects of sensor-data learning: one combines self-supervised encoder training with task-specific perception training, while the other uses paired representations to guide feature learning. His work is associated with Five AI Limited.

Collaborations

Sharma’s listed coworkers include John Redford and Puneet Dokania.

Conclusion

Anuj Sharma’s patent portfolio highlights methods for learning informative sensor-data features through self-supervised training and their use in perception systems.