New project: asl letters

Author: lukas

keras-sign/perceptron.py

@@ -0,0 +1,41 @@
+# A very simple perceptron for classifying american sign language letters
+
+import signdata
+from string import ascii_lowercase
+import numpy as np
+from keras.models import Sequential
+from keras.layers import Dense, Flatten, Dropout, BatchNormalization
+from keras.utils import np_utils
+import wandb
+from wandb.keras import WandbCallback
+
+
+# logging code
+run = wandb.init()
+config = run.config
+
+# load data
+(X_test, y_test) = signdata.load_test_data()
+(X_train, y_train) = signdata.load_train_data()
+
+img_width = X_test.shape[1]
+img_height = X_test.shape[2]
+
+# one hot encode outputs
+y_train = np_utils.to_categorical(y_train)
+y_test = np_utils.to_categorical(y_test)
+
+num_classes = y_train.shape[1]
+
+# you may want to normalize the data here..
+
+# create model
+model=Sequential()
+model.add(Flatten(input_shape=(img_width, img_height)))
+model.add(Dense(num_classes, activation='softmax'))
+model.compile(loss='categorical_crossentropy', optimizer='adam',
+                metrics=['accuracy'])
+
+# Fit the model
+model.fit(X_train, y_train, epochs=10, validation_data=(X_test, y_test),
+                    callbacks=[WandbCallback(data_type="image", labels=ascii_lowercase)])

keras-sign/signdata.py

@@ -0,0 +1,24 @@
+import os.path
+import numpy as np
+import pandas as pd
+import wandb
+
+if not os.path.isfile('sign_mnist_train.csv'):
+    raise ("Can't find data file, please go to https://www.kaggle.com/datamunge/sign-language-mnist and download sign-language-mnist.zip and then unzip in the local directory")
+
+def load_train_data():
+    df=pd.read_csv('sign_mnist_train.csv')
+    X = df.values[:,1:].reshape(-1,28,28)
+    y = df.values[:,0]
+    return X, y
+
+def load_test_data():
+    df=pd.read_csv('sign_mnist_test.csv')
+    X = df.values[:,1:].reshape(-1,28,28)
+    y = df.values[:,0]
+    return X, y
+
+X,y = load_train_data()
+test_X, test_y = load_test_data()
+print(X.shape)
+print(y.shape)