Models

RaulNet Models

RaulNet is a family of CNN-based models for decoding hand kinematics from high-density EMG signals.

V17 (Latest)

The V17 model is the latest version with lazy configuration and TorchScript support.

RaulNetV17(learning_rate, ...[, ...])

Model for decoding kinematics from EMG data.

V16

The V16 model was used in the MyoGestic paper (Simpetru et al., 2024).

RaulNetV16(learning_rate, ...[, ...])

Model definition used in Sîmpetru et al. [1]_.

Components

Activation Functions

Custom learnable activation functions for neural networks.

PSerf([gamma, sigma, stabilisation_term])	PSerf activation function from Biswas et al.
SAU([alpha, n])	SAU activation function from Biswas et al.
SMU([alpha, mu])	SMU activation function from Biswas et al.

Loss Functions

Custom loss functions for kinematics prediction.

EuclideanDistance([n_joints, n_dims])

Euclidean distance loss for 3D joint positions.

Utilities

Utility modules for model building.

WeightedSum([alpha])

Learnable weighted sum of two tensors.