kaira.models.image.DeepJSCCFeedbackEncoder

Inheritance diagram for DeepJSCCFeedbackEncoder

class kaira.models.image.DeepJSCCFeedbackEncoder(conv_depth: int, *args: Any, **kwargs: Any)

Bases: BaseModel

Encoder network for DeepJSCC with Feedback [Kurka and Gündüz, 2020].

This encoder compresses the input image into a latent representation that can be transmitted through a noisy channel. The architecture uses a series of convolutional layers with GDN activations to efficiently encode visual information.

Parameters:

conv_depth (int) – Depth of the output convolutional features, which determines the channel bandwidth usage.

Methods

__init__	Initialize the DeepJSCCFeedbackEncoder.
forward	Forward pass through the encoder.

__init__(conv_depth: int, *args: Any, **kwargs: Any)

Initialize the DeepJSCCFeedbackEncoder.

Parameters:

• conv_depth (int) – Depth of the output convolutional features.

• *args – Variable positional arguments passed to the base class.

• **kwargs – Variable keyword arguments passed to the base class.

forward(x: Tensor, *args: Any, **kwargs: Any) → Tensor

Forward pass through the encoder.

Parameters:

• x (torch.Tensor) – Input image tensor of shape [B, C, H, W].

• *args – Additional positional arguments (passed to internal layers).

• **kwargs – Additional keyword arguments (passed to internal layers).

Returns:

Encoded representation ready for channel transmission.

Return type:

torch.Tensor