Hi on the past days I post Using layers of Pretrained model and Concatenate some additional layers, with your help and reading pytorch doc, I create my model, my goal is to implement " Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields" by Zhe Chaon et al, as a mention before I’m newbie my only experience with PyTorch was train two image classification models, so any guidance would be awesome because I kind loss and I don’t where or how to start with two topics the use COCO dataset keypoints (pose estimation) and how to mount this on a dataloader, sorry for the long post.
The model " Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields" use the first ten layers of VGG19 that’s done and also I add the next layers, this model have two outputs of size (1,19,46,46) and (1,38,46,46), on the GitHub of this work are a way to transfer the dataset to LMDB to use this as input for training but on PyTorch I have not idea, I just found that Pytorch have a module for COCO dataset but is only for object detection and image caption, I was looking for something on GitHub but honestly I feel kind of loss, so any guidance, tip is more than welcome, really thanks for your time and consideration.
1 Like