Liang-Chieh Chen

Title: Innovations of Liang-Chieh Chen

Introduction

Liang-Chieh Chen is a prominent inventor based in Los Angeles, CA. He has made significant contributions to the field of technology, particularly in video segmentation and neural network architectures. With a total of 6 patents, his work has garnered attention in various industries.

Latest Patents

One of his latest patents focuses on implementing video segmentation. This patent describes techniques for dividing a video into multiple clips, each containing several frames. Axial-trajectory attention is applied to these clips by a first sub-model, generating clip features. A transformer decoder then creates a set of object queries based on these features. The process includes refining these queries through trajectory attention, ultimately producing video-level segmentation results.

Another notable patent involves highly efficient convolutional neural networks. This innovation presents new, more efficient neural network architectures. It includes a linear bottleneck layer positioned before and/or after convolutional layers, such as depthwise separable convolutional layers. Additionally, the architecture features inverted residual blocks with thin bottleneck layers for input and output, while an intermediate layer represents an expanded form. This design enhances the efficiency and effectiveness of neural networks.

Career Highlights

Liang-Chieh Chen has worked with notable companies, including Google Inc. and Lemon Inc. His experience in these organizations has contributed to his expertise in developing cutting-edge technologies.

Collaborations

Throughout his career, he has collaborated with esteemed colleagues such as Gerhard Florian Schroff and Hartwig Adam. These partnerships have further enriched his work and innovations.

Conclusion

Liang-Chieh Chen's contributions to technology through his patents and collaborations highlight his role as a leading inventor in the field. His work continues to influence advancements in video processing and neural networks.