USV-based Embedded Obstacle Segmentation Normalization

[Mateus Raitz]

According to the  USV-based Embedded Obstacle Segmentation challenge guidelines,the input format of the model is specified as:

• A single RGB image (scaled to [0-1] range before normalization) normalized by ImageNet mean and standard deviation:
  • mean = [0.485, 0.456, 0.406]
  • std = [0.229, 0.224, 0.225]
• Input shape of the exported model must be 1x3x384x768 (N, C, H, W), while it can be arbitrary during the training.

Does the normalization has to be mean = [0.485, 0.456, 0.406]

std = [0.229, 0.224, 0.225] for the mean and std values or is that just an example?  

Best regards,

Mateus Raitz.

[Matija Tersek]

Hi,

Yes, we currently feed images normalized using ImageNet values for evaluation. You can use different values for training if you want, but the model should then include a transformation which maps normalized input images to your range.

Best,

Matija

Luxonis.com | DepthAI Documentation | Store

Want early access to new products? Join our free Innovation Lab to get early access to hardware and software products. You will also get the opportunity to shape core software solutions built by our development team for your use.

--
Sie erhalten diese Nachricht, weil Sie in Google Groups E-Mails von der Gruppe "MaCVi Support" abonniert haben.
Wenn Sie sich von dieser Gruppe abmelden und keine E-Mails mehr von dieser Gruppe erhalten möchten, senden Sie eine E-Mail an macvi-suppor...@googlegroups.com.
Diese Diskussion finden Sie unter https://groups.google.com/d/msgid/macvi-support/0fcd0cc5-1312-4d18-93bb-076e16de2ecdn%40googlegroups.com.
Weitere Optionen finden Sie unter https://groups.google.com/d/optout.

[Mateus Raitz]

Thanks!