Indian-Sign-Language-Translation-CNN

This was trained from my dataset: Kaggle | Indian Sign Language Translation Letters n Digits

This dataset has 10 digits and 23 out of the 26 letters of the alphabet.

A CNN was trained on this dataset and an accuracy of 98.14% was observed on unseen test dataset.

Model

Results

Dataset Samples

Few classes of the dataset had images similar to these:

Execution Details

The model was trained over Kaggle Kernels using NVidia K80 GPU with 16 GB RAM.

Links

• Kaggel Kernel | Kartik Shenoy | Silatra Digits CNN