ASL_Gesture_CSE535

An online Application Service that accepts Human Pose Skeletal key points of a sign video and return the label of the sign as a JSON Response. The Key points are generated using TensorFlow’s Pose Net. A detailed version of the problem can be found here: ASL_Detect_Problem.pdf

Group Members

• Reet Chatterjee
• Semira Chung
• Baani Khurana
• Rajat Singh

Models

• SVM Linear Regression (Semira)
• Logistic Regression (Baani)
• Random Forest (Rajat)
• SGD Classifier (Reet)

Built With

• Python
• Flask
• AWS EC2
• Nginx
• Gunicorn

Files

• app.py
  • Uses Python Flask to run the app which is running on a server using an API call. Calls main.py to run the project.
• main.py
  • Goes through each and every training data containing JSON objects (book, car, gift, movie, sell, total) and parses them to train all four models.
  • For testing purposes and computing the accuracy, the training data is split into 80% training and 20% test.
  • In order to increase accuracy, unncessary columns are disregarded.
• svm_linear.py
  • A Python file that holds a function called svm_linear(train, test, trainLabels) which uses SVM Linear Regression to predict labels using training and test data.
  • svm is imported from a Python library called sklearn to perform SVM Linear Regression.
• logistic_regression.py
  • A Python file that holds a function called logistic_regression(train, test, trainLabels) which uses Logistic Regression to predict labels using training and test data.
  • LogisticRegression is imported from a Python library called sklearn.linear_model to perform Logistic Regression.
• random_forest.py
  • A Python file that holds a function called random_forest(train, test, trainLabels) which uses the Random Forest model to predict labels using training and test data.
  • RandomForestClassifier is imported from a Python library called sklearn.ensemble to perform the Random Forest functionality.
• sgd_classifier.py
  • A Python file that holds a function called sgd_classifier(train, test, trainLabels) which uses SGD Classifier to predict labels using training and test data.
  • SGDClassifier is imported from a Python library called sklearn.linear_model to perform the SGD Classifier.

NOTE: svm_linear.py, logistic_regression.py, random_forest.py, and sgd_classifier.py use pickle to save the models so that they can be used retrive information more efficiently and quickly instead of re-training the models everytime the application is run.

API Endpoints

• / [Example: http://18.191.155.202/]

  • The Home page displaying the available Endpoints

• /train/ [Example: http://18.191.155.202/train/]

  • The Endpoint is used to call the the function to train four models on the training data. The function also save the models in the server for predictions.

• /pred/ [Example: http://18.191.155.202/pred/]

Note: These endpoints might not work as the free AWS service might have expired.

Request in cURL

curl -X POST
http://18.191.155.202/pred/
-H 'Content-Type: application/json'
-H 'cache-control: no-cache'
-d 'json text'