robot-perception-group / AirPose

This repository contains the code of AirPose, our multi-view fusion network for Human Pose and Shape Estimation method
MIT License
52 stars 6 forks source link

Configuration of bounding box in the synthetic dataset. #3

Open Samuel3Shin opened 1 year ago

Samuel3Shin commented 1 year ago

Hi!

Could you tell me, what is the configuration of the bounding box values in the synthetic dataset?

I'm trying to plot the bounding box of your dataset. But the range of bounding box values is above the image size. How can I interpret this??

Let's say I loaded the 'rp_yumiko_posed_001_0.pkl' dataset in the db variable. I'm trying to plot the bounding box in the image "/copenet_synthetic/data/rp_yumiko_posed_001/MyCamera0_0.png"

Then I need to plot the bounding box with db['bb0'] values. But the values are bb0:<class 'list'>, data: [array([485, 258]), array([614, 436])] which is above the image size.

I saw some processing about the bounding box in the getitem() function aerialpeople.py. But I want to get the real bounding box value for the original image, not for a cropped image or something.

helloyuning commented 5 months ago

Hi!

Could you tell me, what is the configuration of the bounding box values in the synthetic dataset?

I'm trying to plot the bounding box of your dataset. But the range of bounding box values is above the image size. How can I interpret this??

Let's say I loaded the 'rp_yumiko_posed_001_0.pkl' dataset in the db variable. I'm trying to plot the bounding box in the image "/copenet_synthetic/data/rp_yumiko_posed_001/MyCamera0_0.png"

Then I need to plot the bounding box with db['bb0'] values. But the values are bb0:<class 'list'>, data: [array([485, 258]), array([614, 436])] which is above the image size.

I saw some processing about the bounding box in the getitem() function aerialpeople.py. But I want to get the real bounding box value for the original image, not for a cropped image or something.

Hello, DId you repoduce the paper successfully?

nitin-ppnp commented 5 months ago

The bounding box values are for the full-sized images, not the cropped ones. The dataset does not return the full-sized images. It loads the full images, crop and scale them before returning.