predict.py

import cv2
from keras.models import model_from_json
import keras
import numpy as np
import pandas as pd
import pyautogui
import time
import operator
import sys, os
# Loading the model
json_file = open("gesture-model.json", "r")
model_json = json_file.read()
json_file.close()
loaded_model = model_from_json(model_json)
# load weights into new model
loaded_model.load_weights("gesture-model.h5")
print("Loaded model from disk")
final_label = ""
action=""
vid = cv2.VideoCapture(0)
while (vid.isOpened()):

    ret,frame = vid.read()
    if ret:
            frame = cv2.flip(frame, 1)

            # Got this from collect-data.py
            # Coordinates of the ROI
            x1 = int(0.5*frame.shape[1])
            y1 = 10
            x2 = frame.shape[1]-10
            y2 = int(0.5*frame.shape[1])
            # Drawing the ROI
            # The increment/decrement by 1 is to compensate for the bounding box
            cv2.rectangle(frame, (x1-1, y1-1), (x2+1, y2+1), (255,0,0),3)
            # Extracting the ROI
            roi = frame[y1:y2, x1:x2]

            # Resizing the ROI so it can be fed to the model for prediction
            roi = cv2.resize(roi, (120, 120))
            roi = cv2.cvtColor(roi, cv2.COLOR_BGR2GRAY)
            _, test_image = cv2.threshold(roi, 180, 255, cv2.THRESH_BINARY)
            cv2.imshow("Test Image", test_image)
            result = loaded_model.predict(test_image.reshape(1, 120, 120, 1))
            prediction = {'palm': result[0][0],
                          'fist': result[0][1],
                          'thumbs-up': result[0][2],
                          'thumbs-down': result[0][3],
                          'index-right': result[0][4],
                          'index-left': result[0][5],
                          'no-gesture':result[0][6]}
                # Sorting based on top prediction
            prediction = sorted(prediction.items(), key=operator.itemgetter(1), reverse=True)

            if(prediction[0][0] == 'palm'):
                final_label = 'palm'
                action = "PLAY/PAUSE"
                pyautogui.press('playpause', presses=1)
            elif (prediction[0][0] == 'fist'):
                final_label = 'fist'
                action = "MUTE"
                pyautogui.press('volumemute', presses=1)
            elif (prediction[0][0] == 'thumbs-up'):
                final_label = "thumbs-up"
                action = "VOLUME UP"
                pyautogui.press('volumeup', presses=1)
            elif (prediction[0][0] == "thumbs-down"):
                final_label = "thumbs-down"
                action = "VOLUME DOWN"
                pyautogui.press('volumedown', presses=1)
            elif (prediction[0][0] == "index-right"):
                final_label = "index-right"
                action = "FORWARD"
                pyautogui.press('nexttrack', presses=1)
            elif (prediction[0][0] == "index-left"):
                final_label = "index-left"
                action = "REWIND"
                pyautogui.press('prevtrack', presses=1)
            elif (prediction[0][0] == "no-gesture"):
                final_label = "no-gesture"
                action = "NO-ACTION"
            text1= "Gesture: {}".format(final_label)
            text2= "Action:{}".format(action)

            cv2.putText(frame, text1 , (10, 120), cv2.FONT_HERSHEY_COMPLEX_SMALL, 1, (255,0,0), 1)
            cv2.putText(frame, text2 , (10, 220), cv2.FONT_HERSHEY_COMPLEX_SMALL, 1, (255,0,0), 1)
            cv2.imshow("Hand Gesture Recognition", frame)

    if cv2.waitKey(1) & 0xFF == ord('q'):
        break
vid.release()
cv2.destroyAllWindows()