kalyani94/UI_02

1import cv22import numpy as np3 4import tensorflow as tf5#from sklearn.metrics import confusion_matrix6import itertools7import os, glob8from tqdm import tqdm9#from efficientnet.tfkeras import EfficientNetB410 11import tensorflow as tf12from tensorflow.keras.applications.resnet50 import preprocess_input, decode_predictions13from tensorflow.keras.preprocessing import image14from tensorflow.keras.utils import img_to_array, array_to_img15# Helper libraries16import numpy as np17import matplotlib.pyplot as plt18print(tf.__version__)19 20import pandas as pd21import numpy as np22import os23 24import tensorflow as tf25from tensorflow import keras26from tensorflow.keras.preprocessing.image import ImageDataGenerator27from sklearn.preprocessing import LabelBinarizer28 29from IPython.display import clear_output30import warnings31warnings.filterwarnings('ignore')32 33import cv234import gradio as gr35 36 37 38 39 40 41 42labels =['Abuse','Arrest','Arson','Assault','Burglary','Explosion','Fighting',"Normal",'RoadAccidents','Robbery','Shooting','Shoplifting','Stealing','Vandalism']43 44model = keras.models.load_model("classifier.h5")45 46def videoToFrames(video):47 48    # Read the video from specified path 49    cam = cv2.VideoCapture(video) 50 51    '''try: 52        53        # creating a folder named data 54        if not os.path.exists('/home/shubham/__New-D__/VITA/Project/redundant/data/Abuse'): 55            os.makedirs('/home/shubham/__New-D__/VITA/Project/redundant/data/Abuse') 56 57    # if not created then raise error 58    except OSError: 59        print ('Error: Creating directory of data') 60    '''61    62    # frame 63    currentframe = 164    while(True): 65        66        # reading from frame 67        ret,frame = cam.read() 68        69 70        if ret: 71            # if video is still left continue creating images 72            #name = '/home/shubham/__New-D__/VITA/Project/redundant/data/Abuse/frame' + str(currentframe) + '.jpg'73            #print ('Creating...' + name) 74 75            # writing the extracted images 76            77            #cv2.imwrite(name, frame) 78 79            # increasing counter so that it will 80            # show how many frames are created 81            currentframe += 182        else: 83            break84 85    # Release all space and windows once done 86    cam.release() 87    cv2.destroyAllWindows() 88    89    return currentframe90    91def make_average_predictions(video_file_path, predictions_frames_count):92    93    confidences={}94    95    number_of_classes = 1496     97        # Initializing the Numpy array which will store Prediction Probabilities98    predicted_labels_probabilities_np = np.zeros((predictions_frames_count, number_of_classes), dtype = np.float)99 100        # Reading the Video File using the VideoCapture Object101    video_reader = cv2.VideoCapture(video_file_path)102    103        #print(video_reader)104 105        # Getting The Total Frames present in the video106 107    video_frames_count = int(video_reader.get(cv2.CAP_PROP_FRAME_COUNT))108 109        #print(video_frames_count)110 111        # Calculating The Number of Frames to skip Before reading a frame112 113    skip_frames_window = video_frames_count // predictions_frames_count114    115        #print(skip_frames_window)116 117 118 119    for frame_counter in range(predictions_frames_count):120 121        122            # Setting Frame Position123 124        video_reader.set(cv2.CAP_PROP_POS_FRAMES, frame_counter * skip_frames_window)125 126 127 128            # Reading The Frame129 130        _ , frame = video_reader.read()131        132 133 134        image_height, image_width = 64, 64135        136 137            # Resize the Frame to fixed Dimensions138 139        resized_frame = cv2.resize(frame, (image_height, image_width))140 141         142 143            # Normalize the resized frame by dividing it with 255 so that each pixel value then lies between 0 and 1144 145        normalized_frame = resized_frame / 255146 147 148 149            # Passing the Image Normalized Frame to the model and receiving Predicted Probabilities.150 151        predicted_labels_probabilities = model.predict(np.expand_dims(normalized_frame, axis = 0))[0]152 153 154 155            # Appending predicted label probabilities to the deque object156 157        predicted_labels_probabilities_np[frame_counter] = predicted_labels_probabilities158 159 160 161    # Calculating Average of Predicted Labels Probabilities Column Wise162 163    predicted_labels_probabilities_averaged = predicted_labels_probabilities_np.mean(axis = 0)164 165 166 167    # Sorting the Averaged Predicted Labels Probabilities168 169    predicted_labels_probabilities_averaged_sorted_indexes = np.argsort(predicted_labels_probabilities_averaged)[::-1]170 171    predicted_labels_probabilities_averaged_sorted_indexes = predicted_labels_probabilities_averaged_sorted_indexes[:3]172 173    # Iterating Over All Averaged Predicted Label Probabilities174 175    for predicted_label in predicted_labels_probabilities_averaged_sorted_indexes:176 177 178 179        # Accessing The Class Name using predicted label.180 181        predicted_class_name = labels[predicted_label]182 183 184 185        # Accessing The Averaged Probability using predicted label.186 187        predicted_probability = predicted_labels_probabilities_averaged[predicted_label]188 189 190 191        #print(f"CLASS NAME: {predicted_class_name}   AVERAGED PROBABILITY: {(predicted_probability*100):.2}")192        193        confidences[predicted_class_name]=predicted_probability 194        195        196     197 198    # Closing the VideoCapture Object and releasing all resources held by it.199 200    video_reader.release()201    202    return confidences203    204    205def classify_video(video):206 207    framecount = videoToFrames(video)208    confidences = make_average_predictions(video, framecount)209    210    return confidences211    #return confidences212 213demo = gr.Interface(classify_video, 214                    inputs=gr.Video(), 215                    outputs=gr.outputs.Label(),  216                    cache_examples=True)217 218if __name__ == "__main__":219    demo.launch(share=False)220 221 222 223