Simple linear classification model for the CIFAR-10 dataset
A simple linear classification model for the CIFAR-10 dataset with a single fully connected layer consisting of 3072 input channels and 10 output channels.
Model inputs
Image samples in the CIFAR-10 dataset with the following transformations pre-applied.
- Resize each image to 32 by 32 pixels
- Rescale each channel from integer in range
[0, 255]to float in range[0, 1] - Re-order the dimensions from
(height, width, channels)to(channels, height, width)
Model outputs
10 raw logits per image corresponding to the 10 categories in the CIFAR-10 dataset
License
Apache 2.0