Skip to content

abinshoby/DivisiveNet

Repository files navigation

American Sign Language Classification using a Novel CNN Architecture - DivisiveNet

Description

  • Four models - AlexNet, ResNet-18, MobileNetV2 and DivisiveNet are trained for ASL classification.
  • A new CNN architecture - DivisiveNet is developed.
    • Divides the input image continuously until a single dimensional vector is obtained.
    • The single dimesnional vector is further divided using a series of dense layers until 36 output units are produced.
    • Uses batch normalisation, Leaky ReLU after each convolutions and dense layer.
    • Log softmax is used for classification output.
  • All the layers except the final dense layer in the SOTA models are freezed for transfer learning.

Dataset

Code

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published