Recognize_Gesture.py

# coding: utf-8

# In[70]:


import cv2
import numpy as np
from keras.models import load_model


# In[71]:


model = load_model('Gesture_Recognize.h5')


# In[75]:


gesture = {
    0: "You",
    1: "I",
    2: "",
    3: "Love"
}


# In[73]:


def predict(hand):
    img = cv2.resize(hand, (50,50) )
    img = np.array(img)
    img = img.reshape( (1,50,50,1) )
    img = img/255.0
    res = model.predict( img )
    max_ind = res.argmax()
    return gesture[ max_ind ]


# In[85]:


cv2.namedWindow("image", cv2.WINDOW_NORMAL)
cv2.namedWindow("hand", cv2.WINDOW_NORMAL)


vc = cv2.VideoCapture(0)
rval, frame = vc.read()


image_x = 350
image_y = 125
image_w = 200
image_h = 200

old_pred_text = ""
pred_text = ""
count_frames = 0
tot_string = ""

while True:
    
    if frame is not None: 
        
        frame = cv2.flip(frame, 1)
        frame = cv2.resize( frame, (640,480) )
        
        cv2.rectangle(frame, (image_x,image_y), (image_x + image_w,image_y + image_h), (0,255,0), 2)
        
        hand = frame[image_y:image_y+image_h, image_x:image_x+image_w]
        hand = cv2.cvtColor(hand, cv2.COLOR_BGR2GRAY)
        blur = cv2.GaussianBlur(hand, (11,11), 0)
        blur = cv2.medianBlur(blur, 15)
        thresh = cv2.threshold(blur,210,255,cv2.THRESH_BINARY+cv2.THRESH_OTSU)[1]
        thresh = cv2.bitwise_not(thresh)
        
        old_pred_text = pred_text
        
        pred_text = predict(thresh)
        
        if old_pred_text == pred_text:
            count_frames += 1
        else:
            count_frames = 0
        
        
        blackboard = np.zeros(frame.shape, dtype=np.uint8)
        cv2.putText(blackboard, "Predicted text - ", (30, 40), cv2.FONT_HERSHEY_TRIPLEX, 1, (255, 255, 0))
        if count_frames > 20 and pred_text != "":
            tot_string += pred_text + " "
            count_frames = 0
            
        cv2.putText(blackboard, tot_string, (30, 80), cv2.FONT_HERSHEY_TRIPLEX, 1, (255, 255, 127))
        res = np.hstack((frame, blackboard))
        
        cv2.imshow("image", res)
        cv2.imshow("hand", thresh)
        
    rval, frame = vc.read()
    keypress = cv2.waitKey(1)
    
    if keypress == ord('q'):
        break

vc.release()
cv2.destroyAllWindows()
cv2.waitKey(1)