search

karfly / learnable-triangulation-pytorch

This repository is an official PyTorch implementation of the paper "Learnable Triangulation of Human Pose" (ICCV 2019, oral). Proposed method archives state-of-the-art results in multi-view 3D human pose estimation!
MIT License
1.1k stars 181 forks source link

How you prepare the bounding box when transfeering from CMU dataset to Human3.6m? #43

Closed mengmenm closed 4 years ago

mengmenm commented 4 years ago

Amazing jobs, guys.

I am interested in the transformation part mentioned in the paper. And very curious about the bounding box you used when transferring the pre-trained CMU model to the Human3.6m dataset. So which bbox you use when transfer: the original bbox provided by human3.6m or the modified human3.6m bbox which matches the pre-trained model(CMU dataset)?

karfly commented 4 years ago

Hi, @mengmenm. Thank you for a good question. During training on CMU Panoptic dataset we used Mask R-CNN bboxes of humans. When we transfer out models from CMU Panoptic to Human3.6M we use bboxes provided by Human3.6M dataset (GT bboxes). We found out that our models don't overfit to the type of the bbox.

mengmenm commented 4 years ago

hi, @karfly thank you so much.