search

hellojialee / Improved-Body-Parts

Simple Pose: Rethinking and Improving a Bottom-up Approach for Multi-Person Pose Estimation
https://arxiv.org/abs/1911.10529
258 stars 42 forks source link

Where is the coco_train_dataset512.h5 file? #22

Closed FanShuixing closed 4 years ago

FanShuixing commented 4 years ago

Hi,Thanks for your great work.I have a question about the coco_train_dataset512.h5 file which is need to train model.Where can I download it? image

hellojialee commented 4 years ago

Hi friend, coco_train_dataset512.h5 is generated by the script data/coco_masks_hdf5.py. Alternatively, you can rewrite the data loader part reading from the raw images directly.

Sunstin commented 4 years ago

Hi,what is the file coco_train_dataset384.h5 in code coco_masks_hdf5.py? Does it need to download or other way to have it?

Sunstin commented 4 years ago

after I put the coco data under the file, code is successfully running now. But l want to konw why, so i'm reading the code.Why you choose to load the data by create h5 file? @hellojialee Thanks

hellojialee commented 4 years ago

It is actually a legacy manner borrower from other referred repo (keras/tf). Using the hdf5 file can accelerate the data preparation, and we can read and write the hdf5 file in a multi-process in a dataloader Class. I rewrite the dataloader part by reading the raw image data and still achieve high I/O efficiency and GPU utilization. You can refer to other released repo. And I will post this part in the future.

Sunstin commented 4 years ago

Thank you very much for your reply~~

stale[bot] commented 4 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.