Haozhou Zhang
About Haozhou Zhang
Haozhou Zhang is a Computer Vision Research Intern with extensive academic and practical experience in computer science and machine learning. He has contributed to significant advancements in optical flow estimation and drone navigation projects during his internships at various organizations.
Work at Inria
Haozhou Zhang has been working at Inria as a Computer Vision Research Intern since 2019. His role involves research and development in the field of computer vision, particularly focusing on algorithms that enhance image processing. He has contributed to the CLARA project, which is funded by the National Agency for Research, and aims to improve drone navigation in forest environments. His work includes utilizing frameworks such as Caffe and PyTorch for model development and testing.
Education and Expertise
Haozhou Zhang studied at Université de Technologie de Compiègne (UTC), where he earned an Engineer's degree in Computer Science from 2016 to 2019 and a Master's degree in Machine Learning and Optimization of Complex Systems from 2018 to 2019. He also holds a Bachelor's degree in Information Engineering from Shanghai University, which he completed from 2013 to 2017. His academic background provides a strong foundation in computer science and machine learning.
Background
Haozhou Zhang began his academic journey at Shanghai University, where he obtained a Bachelor's degree in Information Engineering. Following this, he pursued further studies at Université de Technologie de Compiègne, achieving both an Engineer's degree and a Master's degree in relevant fields. His internships at Bosch China and herrmann international europe provided him with practical experience in data development and web development, enhancing his technical skills.
Achievements
During his internship at Inria, Haozhou Zhang developed a method to generate 20 equirectangular datasets with ground truth optical flow using Blender's Python API. He implemented a model that improved performance on fine structures and small details by reducing pixels with endpoint error greater than 3 by 10% to 30%. Additionally, he achieved significant enhancements in optical flow estimation, reducing angular error by approximately 0.1 and endpoint error by about 1.