search

hygxy / ma_densefusion

0 stars 0 forks source link

Transferring labelfusion-generated data with multi-object labels #6

Closed WinglokWong616 closed 2 years ago

WinglokWong616 commented 2 years ago

Hi! Me again. It seems that your generatedata.py is meant for data with single object label. Have you ever tried trasferring multi-object labels in one single image?

hygxy commented 2 years ago

this repo doesn't contain a file called "generatedata.py". Do you mean "generategt.py"? No, i haven't tried, but it seems easy, what's your difficulity?

WinglokWong616 commented 2 years ago

this repo doesn't contain a file called "generatedata.py". Do you mean "generategt.py"? No, i haven't tried, but it seems easy, what's your difficulity?

I meant generatedataset.py, just wanna make sure your original code is designed for single object label. But it's okay, problem solved. Thx. Just out of curiosity, I wonder why you chose to transfer your Labelfusion data to Linemod format and train with Linemod mode. 'Cause I found it takes way less steps to transfer mine to YCB format and train with YCB mode. I think they share more similarity.

hygxy commented 2 years ago

Ok, glad to hear that the problem is solved.

The reason why I chose Linemod mode to train is simply that our customer(who provides funding) thinks it's a better choice.

On Tue, 22 Mar 2022 at 16:21, WinglokWong616 @.***> wrote:

this repo doesn't contain a file called "generatedata.py". Do you mean "generategt.py"? No, i haven't tried, but it seems easy, what's your difficulity?

I meant generatedataset.py, just wanna make sure your original code is designed for single object label. But it's okay, problem solved. Thx. Just out of curiosity, I wonder why you chose to transfer your Labelfusion data to Linemod format and train with Linemod mode. 'Cause I found it takes way less steps to transfer mine to YCB format and train with YCB mode. I think they share more similarity.

— Reply to this email directly, view it on GitHub https://github.com/hygxy/ma_densefusion/issues/6#issuecomment-1074867791, or unsubscribe https://github.com/notifications/unsubscribe-auth/AGYMHVESSTFPGJ2ASCWPX2DVBF7J5ANCNFSM5RGP72BA . You are receiving this because you commented.Message ID: @.***>