search

NVlabs / Deep_Object_Pose

Deep Object Pose Estimation (DOPE) – ROS inference (CoRL 2018)
Other
1.03k stars 287 forks source link

Data Generation Question #219

Open xiu-ccc opened 2 years ago

xiu-ccc commented 2 years ago

Hello I am using nvisii_data_gen to generate my own dataset for training. I found that the background of the synthetic dataset is generated by randomly selecting a part of the environment. I want to know how to make the entire environment as the background to generate synthetic data. thank you very much!

TontonTremblay commented 2 years ago

Do you have examples? It is a little hard to imagine what you are describing.

On Fri, Feb 18, 2022 at 07:20 xiu-ccc @.***> wrote:

Hello I am using nvisii_data_gen https://github.com/NVlabs/Deep_Object_Pose/tree/master/scripts/nvisii_data_gen to generate my own dataset for training. I found that the background of the synthetic dataset is generated by randomly selecting a part of the environment. I want to know how to make the entire environment as the background to generate synthetic data. thank you very much!

— Reply to this email directly, view it on GitHub https://github.com/NVlabs/Deep_Object_Pose/issues/219, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABK6JIEYZKHN5GNAUM2ORNDU3ZPVHANCNFSM5OYHKIJA . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

You are receiving this because you are subscribed to this thread.Message ID: @.***>

mintar commented 2 years ago

I believe what @xiu-ccc means is that only part of the background HDRI map is visible in the generated images. @xiu-ccc , is that what you mean?

If yes, I guess there is a misunderstanding. The HDRI maps are projected onto a "dome" that surrounds the whole scene. That means whatever direction the camera is pointing, you will always see only part of the HDRI image. The direction of the camera is randomized every time you start the script.

Or do you mean something else?

xiu-ccc commented 2 years ago

Sorry for my inaccurate description. I have given two example images. 1 00001 Instead of using the HDRI map to generate the environment like nvisii_data_gen, I changed the code to use the image in png format directly. I want to make the background in the generated dataset the entire box in Figure 1, but the final synthesis effect finds that the data synthesized each time can only be a part of this background image.

xiu-ccc commented 2 years ago

I believe what @xiu-ccc means is that only part of the background HDRI map is visible in the generated images. @xiu-ccc , is that what you mean?

If yes, I guess there is a misunderstanding. The HDRI maps are projected onto a "dome" that surrounds the whole scene. That means whatever direction the camera is pointing, you will always see only part of the HDRI image. The direction of the camera is randomized every time you start the script.

Or do you mean something else?

Yes, that's what I want to express, thank you for your help, I understand why I have this problem.