Sign Language Image Classification (SLIC)

Inspiration

We all were interested in image classification and creating a convolutional network, and decided that the ASL dataset presents both a good learning and community challenge.

What it does

• This model attempts to classify individual images into letters of the American Sign Language alphabet. The training data is a large set of sign language images

How we built it

• Using convolutional neural networks on jupyter notebook with python

Challenges we ran into

• Selecting one project out of all the various topics in data science.
• Learning new things.
• Successfully implementing a loss function.
• Translating the image data into a formatted input for the training model.
• Debugging the code
• Understanding how to use pytorch’s CNN model
• Over and underfitting.

Accomplishments that we're proud of

• it works
• _ mostly _

What We Learned

• We had little practical knowledge coming in. We learned almost everything we used today, everything from data cleaning to over-fitting and RELU.

What's next for Sign Language Image Classification (SLIC)

• Complete validation and testing.
• We have neither time or resources to implement SLIC into video formats, this is a future goal.

Built With

• convolutional-neural-networks
• numpy
• python
• pytorch
• scikit-learn