CIFAR CNN

$0.00

A CNN classifier model as a block to be trained on the CIFAR dataset.

Clear
Report Abuse

Report an abuse for product CIFAR CNN

SKU: N/A Category:
Version 0.2.4-x86_64
Input ports
  • Image Batch (perception_msgs/ImageBatch) - The batch of images streamed from a player with a size equals to the batch size property.
  • Labels (perception_msgs/LabelArray) - An array of labels streamed from a player with a size equals to the batch size property.
Output Ports
  • Loss (yonoarc_msgs/Float64) - The loss value of the training each epoch
  • Accuracy (yonoarc_msgs/Float64) - The accuracy percentage of the training each epoch
Properties
  • Batch Size - The batch size expected to be streamed by the player. Normally, you can set it with the same value as the player batch size using a global parameter.
  • Dataset Size - The train split size used to train the model. This is used to calculated the number of mini-batches.
  • Epochs - The number of epochs/ training loops you want to train your model.
  • Learning Rate - The learning rate of the ADAM optimizer
  • Model Mode - Select the model mode that suits your application.
  • Model Name - This is the name of your model which will be used to name your saved .h5 trained model.
  • Model Path - The location that should be used to save your .h5 trained model.
  • Momentum - The momentum or beta 1 of the ADAM optimizer
  • SSH Password - The password of the ssh access of the block
  • Tensorboad Log Directory - The directory to be used to save the logs for Tensorboard.
  • SSH Username - The username of ssh access of the block
References None

Reviews

There are no reviews yet.

Be the first to review “CIFAR CNN”

Your email address will not be published. Required fields are marked *

Questions and Answers

You are not logged in