Inspiration
We chose this classification problem with the long term goal of building an app that will make it easier to learn american sign language in order to build a more inclusive world.
What it does
We are translating images of the american sign language alphabet and numbers into text.
How we built it
We used the pretrained model ResNet50 and added some layers to it.
Challenges we ran into
Since we coded on google colab without the pro version, we were very limited in terms of RAM and GPU, hence we needed our code to be efficient.
Accomplishments that we're proud of
We got a validation accuracy of 97.1% by training our model on the alphabet and the numbers datasets. By only training our model on the numbers dataset, we got 99.73% validation accuracy.
What's next for ASL alphabet and numbers classifier
If we were to continue this project, we would like to build a learning app for the american sign language alphabet and numbers.
Built With
- googlecollab
- python
Log in or sign up for Devpost to join the conversation.