Ashish Patel 🇮🇳’s Post

𝗗𝗮𝘆-𝟰𝟴𝟬 𝗖𝗼𝗺𝗽𝘂𝘁𝗲𝗿 𝗩𝗶𝘀𝗶𝗼𝗻 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 StyleGAN-Human: A Data-Centric Odyssey of Human Generation by Shanghai AI Laboratory, China Follow me for a similar post: Ashish Patel ------------------------------------------------------------------- 𝗜𝗻𝘁𝗲𝗿𝗲𝘀𝘁𝗶𝗻𝗴 𝗙𝗮𝗰𝘁𝘀 : 🔸 This paper is published #arxiv 2022. ------------------------------------------------------------------- 𝗜𝗠𝗣𝗢𝗥𝗧𝗔𝗡𝗖𝗘 👉 Unconditional human image generation is an important task in vision and graphics, which enables various applications in the creative industry.  👉 Existing studies in this field mainly focus on "network engineering" such as designing new components and objective functions.  👉 This work takes a data-centric perspective and investigates multiple critical aspects of "data engineering", which we believe would complement the current practice.  👉 To facilitate a comprehensive study, we collect and annotate a large-scale human image dataset with over 230K samples capturing diverse poses and textures.  👉 Equipped with this large dataset, we rigorously investigate three essential factors in data engineering for StyleGAN-based human generation, namely data size, data distribution, and data alignment.  👉 Extensive experiments reveal several valuable observations w.r.t. these aspects:  👉 1) Large-scale data, more than 40K images, are needed to train a high-fidelity unconditional human generation model with vanilla StyleGAN.  👉 2) A balanced training set helps improve the generation quality with rare face poses compared to the long-tailed counterpart, whereas simply balancing the clothing texture distribution does not effectively bring an improvement.  👉 3) Human GAN models with body centers for alignment outperform models trained using face centers or pelvis points as alignment anchors. In addition, a model zoo and human editing applications are demonstrated to facilitate future research in the community. #computervision #artificialintelligence #deeplearning #datascience #machinelearning #technology #india #data