Deep_Object_Pose icon indicating copy to clipboard operation
Deep_Object_Pose copied to clipboard
verified publisher icon NVlabs

Making the format of the dataset

Open JYX0118 opened this issue 2 years ago • 3 comments

Hello, are there any formatting requirements for creating my own dataset, and I look forward to your reply!

JYX0118 avatar Apr 26 '23 14:04 JYX0118

What do you mean? There is an example json file you can look into if you want to know more. https://github.com/NVlabs/Deep_Object_Pose/blob/master/scripts/nvisii_data_gen/output/output_example/00000.json The training script needs similar data to this.

TontonTremblay avatar Apr 26 '23 16:04 TontonTremblay

Thank you for your answer, I still have the following questions:

  1. If I need to identify my own objects and I have to make my own dataset, which tool is better to use to make the dataset.
  2. What should I pay attention to when training my own dataset, for example, should I use scripts/train.py or scripts/train2/train.py, and do I need to modify other places? 3、When I have trained my own dataset and got the model, which folder should I put the model into? Then launch camera.launch and dope.launch, I know there must be some parameters that need to be modified, but I'm not quite sure which places need to be modified. Looking forward to your answer, thanks again!

JYX0118 avatar Apr 27 '23 02:04 JYX0118

  1. Nvisii right now is the most complete tool you should use, check the script for generating training data.
  2. train2 has an output in the tensorboard, you need to see belief maps being visible.
  3. Follow the instructions on how to use the config file with respect to object sizes.

I hope this helps.

On Wed, Apr 26, 2023 at 7:32 PM JYX @.***> wrote:

Thank you for your answer, I still have the following questions:

  1. If I need to identify my own objects and I have to make my own dataset, which tool is better to use to make the dataset.
  2. What should I pay attention to when training my own dataset, for example, should I use scripts/train.py or scripts/train2/train.py, and do I need to modify other places? 3、When I have trained my own dataset and got the model, which folder should I put the model into? Then launch camera.launch and dope.launch, I know there must be some parameters that need to be modified, but I'm not quite sure which places need to be modified. Looking forward to your answer, thanks again!

— Reply to this email directly, view it on GitHub https://github.com/NVlabs/Deep_Object_Pose/issues/298#issuecomment-1524502407, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABK6JIBRZ5ILRX6F7OIJURLXDHLELANCNFSM6AAAAAAXMRAH4E . You are receiving this because you commented.Message ID: @.***>

TontonTremblay avatar Apr 27 '23 17:04 TontonTremblay