Skip to content

Training a model (Encoder and a Classifier) with a standard decoder capable of compressing image and using the latent representation for classification

License

Notifications You must be signed in to change notification settings

riaddot/Image_comp_classification_CIFAR10

Repository files navigation

Image_comp_classification_CIFAR10

Training a model (Encoder and a Classifier) with a standard decoder capable of compressing image and using the latent representation for classification

In this model we took the weight of the decoder in High Fidelity Generative Image Compression and we trained both the Encoder and the Classifier. with a loss function that combines both the restoration loss and the classificatino loss. We used CIFAR-10 as our dataset.

The global architecture can be shown in the following figure:

image

About

Training a model (Encoder and a Classifier) with a standard decoder capable of compressing image and using the latent representation for classification

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages