first commit

HandGestureRecognize.py

@@ -0,0 +1,259 @@
+from tkinter import messagebox
+from tkinter import *
+from tkinter import simpledialog
+import tkinter
+from tkinter import filedialog
+from tkinter.filedialog import askopenfilename
+import cv2
+import random
+import numpy as np
+from keras.utils.np_utils import to_categorical
+from keras.layers import  MaxPooling2D
+from keras.layers import Dense, Dropout, Activation, Flatten
+from keras.layers import Convolution2D
+from keras.models import Sequential
+from keras.models import model_from_json
+import pickle
+import os
+import imutils
+from gtts import gTTS   
+from playsound import playsound
+import os
+from threading import Thread 
+
+main = tkinter.Tk()
+main.title("Hand geture recognition and voice conversation using CNN")
+main.geometry("1200x600")
+
+global filename
+global classifier
+
+bg = None
+playcount = 0
+
+#names = ['Palm','I','Fist','Fist Moved','Thumbs up','Index','OK','Palm Moved','C','Down']
+names = ['C','Thumbs Down','Fist','I','Ok','Palm','Thumbs up']
+
+def getID(name):
+    index = 0
+    for i in range(len(names)):
+        if names[i] == name:
+            index = i
+            break
+    return index  
+
+
+bgModel = cv2.createBackgroundSubtractorMOG2(0, 50)
+
+def deleteDirectory():
+    filelist = [ f for f in os.listdir('play') if f.endswith(".mp3") ]
+    for f in filelist:
+        os.remove(os.path.join('play', f))
+
+def play(playcount,gesture):
+    class PlayThread(Thread):
+        def __init__(self,playcount,gesture):
+            Thread.__init__(self) 
+            self.gesture = gesture
+            self.playcount = playcount
+
+        def run(self):
+            t1 = gTTS(text=self.gesture, lang='en', slow=False)
+            t1.save("play/"+str(self.playcount)+".mp3")
+            playsound("play/"+str(self.playcount)+".mp3")
+
+
+    newthread = PlayThread(playcount,gesture) 
+    newthread.start()            
+
+def remove_background(frame):
+    fgmask = bgModel.apply(frame, learningRate=0)
+    kernel = np.ones((3, 3), np.uint8)
+    fgmask = cv2.erode(fgmask, kernel, iterations=1)
+    res = cv2.bitwise_and(frame, frame, mask=fgmask)
+    return res
+
+def uploadDataset():
+    global filename
+    global labels
+    labels = []
+    filename = filedialog.askdirectory(initialdir=".")
+    pathlabel.config(text=filename)
+    text.delete('1.0', END)
+    text.insert(END,filename+" loaded\n\n");
+
+
+
+def trainCNN():
+    global classifier
+    text.delete('1.0', END)
+    X_train = np.load('model1/X.txt.npy')
+    Y_train = np.load('model1/Y.txt.npy')
+    text.insert(END,"CNN is training on total images : "+str(len(X_train))+"\n")
+
+    if os.path.exists('model1/model.json'):
+        with open('model1/model.json', "r") as json_file:
+            loaded_model_json = json_file.read()
+            classifier = model_from_json(loaded_model_json)
+        classifier.load_weights("model1/model_weights.h5")
+        classifier._make_predict_function()   
+        print(classifier.summary())
+        f = open('model1/history.pckl', 'rb')
+        data = pickle.load(f)
+        f.close()
+        acc = data['accuracy']
+        accuracy = acc[9] * 100
+        text.insert(END,"CNN Hand Gesture Training Model Prediction Accuracy = "+str(accuracy))
+    else:
+        classifier = Sequential()
+        classifier.add(Convolution2D(32, 3, 3, input_shape = (64, 64, 3), activation = 'relu'))
+        classifier.add(MaxPooling2D(pool_size = (2, 2)))
+        classifier.add(Convolution2D(32, 3, 3, activation = 'relu'))
+        classifier.add(MaxPooling2D(pool_size = (2, 2)))
+        classifier.add(Flatten())
+        classifier.add(Dense(output_dim = 256, activation = 'relu'))
+        classifier.add(Dense(output_dim = 5, activation = 'softmax'))
+        print(classifier.summary())
+        classifier.compile(optimizer = 'adam', loss = 'categorical_crossentropy', metrics = ['accuracy'])
+        hist = classifier.fit(X_train, Y_train, batch_size=16, epochs=10, shuffle=True, verbose=2)
+        classifier.save_weights('model1/model_weights.h5')            
+        model_json = classifier.to_json()
+        with open("model1/model.json", "w") as json_file:
+            json_file.write(model_json)
+        f = open('model1/history.pckl', 'wb')
+        pickle.dump(hist.history, f)
+        f.close()
+        f = open('model1/history.pckl', 'rb')
+        data = pickle.load(f)
+        f.close()
+        acc = data['accuracy']
+        accuracy = acc[9] * 100
+        text.insert(END,"CNN Hand Gesture Training Model Prediction Accuracy = "+str(accuracy))
+
+
+
+def run_avg(image, aWeight):
+    global bg
+    if bg is None:
+        bg = image.copy().astype("float")
+        return
+    cv2.accumulateWeighted(image, bg, aWeight)
+
+def segment(image, threshold=25):
+    global bg
+    diff = cv2.absdiff(bg.astype("uint8"), image)
+    thresholded = cv2.threshold(diff, threshold, 255, cv2.THRESH_BINARY)[1]
+    ( cnts, _) = cv2.findContours(thresholded.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
+    if len(cnts) == 0:
+        return
+    else:
+        segmented = max(cnts, key=cv2.contourArea)
+        return (thresholded, segmented)
+
+
+def webcamPredict():
+    global playcount
+    oldresult = 'none'
+    count = 0
+    fgbg2 = cv2.createBackgroundSubtractorKNN(); 
+    aWeight = 0.5
+    camera = cv2.VideoCapture(0)
+    top, right, bottom, left = 10, 350, 325, 690
+    num_frames = 0
+    while(True):
+        (grabbed, frame) = camera.read()
+        frame = imutils.resize(frame, width=700)
+        frame = cv2.flip(frame, 1)
+        clone = frame.copy()
+        (height, width) = frame.shape[:2]
+        roi = frame[top:bottom, right:left]
+        gray = cv2.cvtColor(roi, cv2.COLOR_BGR2GRAY)
+        gray = cv2.GaussianBlur(gray, (41, 41), 0)
+        if num_frames < 30:
+            run_avg(gray, aWeight)
+        else:
+            temp = gray
+            hand = segment(gray)
+            if hand is not None:
+                (thresholded, segmented) = hand
+                cv2.drawContours(clone, [segmented + (right, top)], -1, (0, 0, 255))
+                #cv2.imwrite("test.jpg",temp)
+                #cv2.imshow("Thesholded", temp)
+                #ret, thresh = cv2.threshold(temp, 150, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)
+                #thresh = cv2.resize(thresh, (64, 64))
+                #thresh = np.array(thresh)
+                #img = np.stack((thresh,)*3, axis=-1)
+                roi = frame[top:bottom, right:left]
+                roi = fgbg2.apply(roi); 
+                cv2.imwrite("test.jpg",roi)
+                #cv2.imwrite("newDataset/Fist/"+str(count)+".png",roi)
+                #count = count + 1
+                #print(count)
+                img = cv2.imread("test.jpg")
+                img = cv2.resize(img, (64, 64))
+                img = img.reshape(1, 64, 64, 3)
+                img = np.array(img, dtype='float32')
+                img /= 255
+                predict = classifier.predict(img)
+                value = np.amax(predict)
+                cl = np.argmax(predict)
+                result = names[np.argmax(predict)]
+                if value >= 0.99:
+                    print(str(value)+" "+str(result))
+                    cv2.putText(clone, 'Gesture Recognize as : '+str(result), (10, 25),  cv2.FONT_HERSHEY_SIMPLEX,0.5, (0, 255, 255), 2)
+                    if oldresult != result:
+                        play(playcount,result)
+                        oldresult = result
+                        playcount = playcount + 1
+                else:
+                    cv2.putText(clone, '', (10, 25),  cv2.FONT_HERSHEY_SIMPLEX,0.5, (0, 255, 255), 2)
+                cv2.imshow("video frame", roi)
+        cv2.rectangle(clone, (left, top), (right, bottom), (0,255,0), 2)
+        num_frames += 1
+        cv2.imshow("Video Feed", clone)
+        keypress = cv2.waitKey(1) & 0xFF
+        if keypress == ord("q"):
+            break
+    camera.release()
+    cv2.destroyAllWindows()    
+
+
+
+
+font = ('times', 16, 'bold')
+title = Label(main, text='Hand geture recognition and voice conversation using CNN',anchor=W, justify=CENTER)
+title.config(bg='yellow4', fg='white')  
+title.config(font=font)           
+title.config(height=3, width=120)       
+title.place(x=0,y=5)
+
+
+font1 = ('times', 13, 'bold')
+upload = Button(main, text="Upload Hand Gesture Dataset", command=uploadDataset)
+upload.place(x=50,y=100)
+upload.config(font=font1)  
+
+pathlabel = Label(main)
+pathlabel.config(bg='yellow4', fg='white')  
+pathlabel.config(font=font1)           
+pathlabel.place(x=50,y=150)
+
+markovButton = Button(main, text="Train CNN with Gesture Images", command=trainCNN)
+markovButton.place(x=50,y=200)
+markovButton.config(font=font1)
+
+predictButton = Button(main, text="Sign Language Recognition from Webcam", command=webcamPredict)
+predictButton.place(x=50,y=250)
+predictButton.config(font=font1)
+
+
+font1 = ('times', 12, 'bold')
+text=Text(main,height=15,width=78)
+scroll=Scrollbar(text)
+text.configure(yscrollcommand=scroll.set)
+text.place(x=450,y=100)
+text.config(font=font1)
+
+deleteDirectory()
+main.config(bg='magenta3')
+main.mainloop()

model1/model.json

@@ -0,0 +1 @@
+{"class_name": "Sequential", "config": {"name": "sequential_1", "layers": [{"class_name": "Conv2D", "config": {"name": "conv2d_1", "trainable": true, "batch_input_shape": [null, 64, 64, 3], "dtype": "float32", "filters": 32, "kernel_size": [3, 3], "strides": [1, 1], "padding": "valid", "data_format": "channels_last", "dilation_rate": [1, 1], "activation": "relu", "use_bias": true, "kernel_initializer": {"class_name": "VarianceScaling", "config": {"scale": 1.0, "mode": "fan_avg", "distribution": "uniform", "seed": null}}, "bias_initializer": {"class_name": "Zeros", "config": {}}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}}, {"class_name": "MaxPooling2D", "config": {"name": "max_pooling2d_1", "trainable": true, "dtype": "float32", "pool_size": [2, 2], "padding": "valid", "strides": [2, 2], "data_format": "channels_last"}}, {"class_name": "Conv2D", "config": {"name": "conv2d_2", "trainable": true, "dtype": "float32", "filters": 32, "kernel_size": [3, 3], "strides": [1, 1], "padding": "valid", "data_format": "channels_last", "dilation_rate": [1, 1], "activation": "relu", "use_bias": true, "kernel_initializer": {"class_name": "VarianceScaling", "config": {"scale": 1.0, "mode": "fan_avg", "distribution": "uniform", "seed": null}}, "bias_initializer": {"class_name": "Zeros", "config": {}}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}}, {"class_name": "MaxPooling2D", "config": {"name": "max_pooling2d_2", "trainable": true, "dtype": "float32", "pool_size": [2, 2], "padding": "valid", "strides": [2, 2], "data_format": "channels_last"}}, {"class_name": "Flatten", "config": {"name": "flatten_1", "trainable": true, "dtype": "float32", "data_format": "channels_last"}}, {"class_name": "Dense", "config": {"name": "dense_1", "trainable": true, "dtype": "float32", "units": 256, "activation": "relu", "use_bias": true, "kernel_initializer": {"class_name": "VarianceScaling", "config": {"scale": 1.0, "mode": "fan_avg", "distribution": "uniform", "seed": null}}, "bias_initializer": {"class_name": "Zeros", "config": {}}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}}, {"class_name": "Dense", "config": {"name": "dense_2", "trainable": true, "dtype": "float32", "units": 7, "activation": "softmax", "use_bias": true, "kernel_initializer": {"class_name": "VarianceScaling", "config": {"scale": 1.0, "mode": "fan_avg", "distribution": "uniform", "seed": null}}, "bias_initializer": {"class_name": "Zeros", "config": {}}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}}]}, "keras_version": "2.3.1", "backend": "tensorflow"}