search

Nicholasli1995 / EvoSkeleton

Official project website for the CVPR 2020 paper (Oral Presentation) "Cascaded deep monocular 3D human pose estimation wth evolutionary training data"
https://arxiv.org/abs/2006.07778
MIT License
333 stars 44 forks source link

Fine-tuning of HRNet #84

Open liujingtingdev opened 5 months ago

liujingtingdev commented 5 months ago

Could you please share your approach for converting the Human3.6M dataset to COCO format for fine-tuning? Would it be possible to provide your image conversion files, the resulting JSON file format, and the joint weights?

Nicholasli1995 commented 5 months ago

Thank you for your interest in this study. I did not convert the H3.6M data format back to COCO. Instead, I used model weights pre-trained on the COCO dataset to directly initialize a model to train on the H3.6M dataset.