Wen Dai

Title: Wen Dai: Innovations in Neural Network Inference and Photogrammetry.

Introduction

Wen Dai is an inventor associated with Nanjing, China. Dai’s patent work spans neural network inference acceleration and photogrammetric quality evaluation. The supplied information lists 3 patents.

Latest Patents

One of Dai’s latest patents, “Method and Apparatus for Accelerating Inference of Neural Network Model, Electronic Device, and Medium,” describes a process for making a neural network model run faster. It converts an operation in a selected network layer to a substitute operation that simulates the target operation, then applies quantization-aware training at a preset bit width. The trained model is converted back to use the target operation, producing an accelerated model. The method can use image, text, or speech training data.

Another patent, “Photogrammetric Ground Control Point Quality Evaluation Method Based on Monte Carlo Test,” presents a way to assess ground control points used in photogrammetry. It uses aerial photographs and ground control points, repeatedly assigns points as controls and checks, and calculates average errors and their standard deviations. A quality coefficient is then determined for each point according to evaluation standards.

Career Highlights

Dai’s listed professional affiliations include Nanjing University of Information Science & Technology and Nanjing Horizon Robotics Integrated Circuit Co., Ltd. The patent topics reflect work across machine learning systems and surveying technology.

Collaborations

The supplied information identifies Bo Wang and Guojie Wang as Dai’s coworkers. Their names are included as professional collaborators, without further details about their specific roles in the listed patents.

Conclusion

Wen Dai’s patent portfolio, as described in the supplied information, brings together techniques for accelerating neural network inference and evaluating photogrammetric ground control points.