KaneAI - World’s First E2E Software Testing Agent.Book a Demo Join TestMu Conference

How to use test_gesture method in uiautomator

Best Python code snippet using uiautomator

model.py

Source:model.py

1import csv2import time3import matplotlib.pyplot as plt4import features5import knn6import sys7FNAME_IN_PREFIX = "../data/2022-03-22"8FNAME_OUT_PREFIX = "model/2022-03-22"9#TNUM = 510#GESTURES = [["LP", "TA", "TLFO", "TIFO", "TLFE", "TIFE", "TIFE", "LFE", "IFE", "TE", "WF", "WE", "FS", "FP", "HO", "HC", "HR"]11GESTURES = ["HO", "HR", "HC", "TIFE"]12DIM_NUM = 613feature_sets = []14def load_feature_sets_from_file():15    fsets = []16    try:17        for g in GESTURES:18            fsets.append(list(csv.reader(open("{}/model_{}_1.csv".format(FNAME_OUT_PREFIX, g.lower())))))19    except Exception as e:20        print("Error loading {}: {}".format(g, e))21        return None22    return fsets23def get_feature_sets(force_remodel=False):24    print("Loading feature sets")25    26    if not force_remodel:27        fsets = load_feature_sets_from_file()28        if fsets:29            print("Loaded feature sets from file successfully")30            return fsets31    fsets = []32    for g in GESTURES:33        pnum = 1 # participant34        print("Modelling {}...".format(g))35        f = []36        for tnum in range(TNUM):37            f += features.get_features(features.import_data("{}/p{}_{}_t{}.csv".format(FNAME_IN_PREFIX, pnum, g.lower(), tnum + 1)), g)38        features.write_features_to_csv("{}/model_{}.csv".format(FNAME_OUT_PREFIX, g.lower(), tnum + 1), f)39        fsets.append(f)40    return load_feature_sets_from_file()41def load_test_points(test_gesture):42    return list(csv.reader(open("{}/model_{}_1.csv".format(FNAME_OUT_PREFIX, test_gesture.lower()))))43# 9s to model everything and write to csv44# response time should be under 300ms45# TODO FIX THIS46#feature_sets = get_feature_sets(force_remodel=False)47def calc_certainty(label, kn):48    return round([p[0] for p in kn].count(label) / len(kn), 2)49feature_sets = []50print("Generating model...")51for g in GESTURES:52    # Generate features53    f = features.get_features(features.import_data("{}/{}.csv".format(FNAME_IN_PREFIX, g.lower())), g.lower(), DIM_NUM)54    55    # Save model56    features.write_features_to_csv("{}/model_{}_{}.csv".format(FNAME_OUT_PREFIX, g.lower(), 1), f)57    feature_sets.append(f)58# TODO: maybe implement weighting59K = 360bar = False61realtime = True62WINDOW_LEN_MS = int(256 / 1260 * 1000)63knn_data = []64for i in range(len(GESTURES)):65    knn.add_training_data(knn_data, feature_sets[i], GESTURES[i])66#for test_gesture in GESTURES:67test_gesture = "test2"68test_points = load_test_points(test_gesture)69start = time.perf_counter()70classification_results = []71classification_bests = [("TIFE", 1)]72classification_filtered = ["TIFE"]73DELAY_MS = 240074#for g in GESTURES:75    #classification_results[g] = 076test_len = 10977if realtime:78    input("Press any key to start real-time classification...")79for i in range(test_len):80    ps = time.perf_counter()81    k_nearest = knn.get_k_nearest(knn_data, test_points[i], K)82    label = knn.calc_mode(k_nearest)83    classification_results.append(k_nearest)84    certainty = calc_certainty(label, k_nearest) 85    if realtime:86        time.sleep((WINDOW_LEN_MS - (time.perf_counter() - ps)) / 1000)87    # filter88    cur_label = classification_filtered[-1]89    accepted = False90    if i >= 2 and certainty >= 1:91        # TIFE is rarely misclassified92        if label == "TIFE":93            accepted = True94        # As gesture recognition is mostly identified throughout the initial transition into the gesture, indentified periods of rest are normal throughout the gesture and can be ignored95        elif label == "HR":96            accepted = False97        # For other gestures, ensure that the previous matches98        elif label != classification_filtered[-1]:99            if (label == classification_bests[-1][0] and classification_bests[-1][1] >= 0.6):100                accepted = True101            else:102                accepted = False103    else:104        accepted = False105    if accepted:106        ms = (i + 1) * WINDOW_LEN_MS107        print("[{}]: {}".format(ms, label))108        classification_filtered.append(label)109    else:110        classification_filtered.append(cur_label)111    classification_bests.append((label, certainty))112    #print("{} ({}%)".format(label, int(certainty * 100)))113    sys.stdout.flush()114end = time.perf_counter()115total_num = len(test_points)116duration_ms = (end - start) * 1000117print("{}ms of input data processed".format(WINDOW_LEN_MS * len(classification_results)))118print("Duration: {}ms for {} test points, {}ms per point".format(duration_ms, len(test_points), duration_ms/len(test_points)))119#print("Accuracy {} (K={}) {}%".format(test_gesture, K, int(classification_results[test_gesture] * 100 / total_num)))120imgname = ""121plt.title("First layer Classifier with K-NN (K={})".format(K))122plt.xticks(fontsize=7)123if bar:124    plt.bar(classification_results.keys(), [x/total_num for x in classification_results.values()])125    imgname = "img/{}_k{}_all.png"126else:127    total_len = WINDOW_LEN_MS * len(test_points)128    mapping = dict([(5, dict([(0.2, 1), (0.4, 4), (0.6, 32), (0.8, 64), (1, 128)])), (3, dict([(0.33, 1), (0.67, 32), (1, 128)]))])129    plt.scatter([x * WINDOW_LEN_MS for x in range(len(classification_bests))], [x[0] for x in classification_bests], s=[mapping[K][x[1]] for x in classification_bests])130    imgname = "img/{}_k{}_bests.png"131plt.savefig("out/first.png")132plt.show()133plt.clf()134plt.title("Second layer Classifier")135plt.xticks(fontsize=7)136plt.scatter([x * WINDOW_LEN_MS for x in range(len(classification_filtered))], classification_filtered)137plt.savefig("out/second.png")138plt.show()139# certaiintty140plt.clf()141plt.title("Certainty")142plt.xticks(fontsize=7)143plt.scatter([x * WINDOW_LEN_MS for x in range(len(classification_bests))], [x[1] for x in classification_bests])144plt.savefig("out/certainty.png")...

main.py

Source:main.py

1import csv2import fnmatch3import os4import cv25import numpy as np6import tensorflow as tf7import handshape_feature_extractor8import frameextractor9# global variables10BASE = os.path.dirname(os.path.abspath(__file__))11IMAGE_MODE = cv2.IMREAD_GRAYSCALE12TRAININGDATA = os.path.join(BASE, 'traindata')13TESTINGDATA = os.path.join(BASE, 'test')14TRAINOUT = os.path.join(BASE, 'trainout')15TESTOUT = os.path.join(BASE, 'testout')16# CNN model17hfe = handshape_feature_extractor.HandShapeFeatureExtractor().get_instance()18table = {19    'Num0': 0, 'num0': 0,20    'Num1': 1, 'num1': 1,21    'Num2': 2, 'num2': 2,22    'Num3': 3, 'num3': 3,23    'Num4': 4, 'num4': 4,24    'Num5': 5, 'num5': 5,25    'Num6': 6, 'num6': 6,26    'Num7': 7, 'num7': 7,27    'Num8': 8, 'num8': 8,28    'Num9': 9, 'num9': 9,29    'FanDown': 10, 'decreasefanspeed': 10,'FanSpeedDown':10,30    'FanOff': 11, 'fanoff': 11,31    'FanOn': 12, 'fanon': 12,32    'FanUp': 13, 'increasefanspeed': 13,'FanSpeedUp': 13,33    'LightOff': 14, 'lightoff': 14,34    'LightOn': 15, 'lighton': 15,35    'SetThermo': 16, 'setthermo': 16, 'Sethermostat':1636}37class Gesture:38    """Holds the data pertaining to gesture like the name,video file name, image file name etc"""39    def __init__(self, name=None, video_file=None, image_file=None, feature_score=None,40                 true_label=float('inf'), predicted_label=float('inf')):41        """42        : name -Gesture name43        : video_file -video file name44        : image_file -image file name45        : feature_score -  extract_feature() score46        : true_label - true label for training data47        : predicted_label -predicted label after cosine similarity48        """49        self.name = name50        self.image_file = image_file51        self.feature_score = feature_score52        self.video_file = video_file53        self.true_label = true_label54        self.predicted_label = predicted_label55        if self.video_file:56            self.video_name = self.video_file.replace('_SaiMadhuriMolleti', '').replace('.mp4', '')57    def is_identified(self):58        return self.predicted_label == self.true_label59def features_test():60    """ extracts the features from each testdata video provided  """61    test_gesture = dict()62    for test_count, test_file in enumerate(fnmatch.filter(os.listdir(TESTINGDATA), '*.mp4')):63        te_gesture = feature_gesture(TESTINGDATA, TESTOUT, test_file, test_count)64        test_gesture[te_gesture.video_name] = te_gesture65    return test_gesture66def features_train():67    """ extracts the features from each train data video provided """68    train_gesture = dict()69    for train_count, train_file in enumerate(fnmatch.filter(os.listdir(TRAININGDATA), '*.mp4')):70        tr_gesture = feature_gesture(TRAININGDATA, TRAINOUT, train_file, train_count)71        train_gesture[tr_gesture.video_name] = tr_gesture72    return train_gesture73def get_gesture_name(file_path):74    name = file_path75    if '-' in file_path:76        name = '-'.join(file_path.split('-')[1:])77    elif '_' in file_path:78        name = file_path.split('_')[0]79    return name.replace(".mp4", "")80def feature_gesture(data_dir, out_dir, video_file, count):81    """Extract frame to a file and calculate its feature vector82    """83    video_path = os.path.join(data_dir, video_file)84    frameextractor.frameExtractor(video_path, out_dir, count)85    image_path = os.path.join(out_dir, '{0:05}.png'.format(count+1))86    gesture_name = get_gesture_name(video_file)87    image_file = os.path.basename(image_path)88    image_array = cv2.imread(image_path, IMAGE_MODE)89    vector = hfe.extract_feature(image_array)90    try:91        # get the known true label for train data92        true_label = table[gesture_name]93    except KeyError:94        # guess the expected true label for test data95        true_label = count % 1796    gesture = Gesture(gesture_name, video_file, image_file,97                      vector, true_label)98    return gesture99def print_result(test_gesture, results_matrix):100    correct = np.sum([1 for x in test_gesture if test_gesture[x].is_identified()])101    total = results_matrix.sum()102    percentage = correct/total*100103    print("Matched {0} out of {1} => {2:0.2f} %".format(correct, total, percentage))104    with open(os.path.join(BASE, 'Results.csv'), mode='w') as results_csv:105        results_writer = csv.writer(results_csv, delimiter=',', quotechar='"', quoting=csv.QUOTE_MINIMAL)106        for i in sorted(test_gesture, key=lambda x: test_gesture[x].video_name):107            results_writer.writerow([test_gesture[i].predicted_label])108def main():109    test_gesture = features_test()110    train_gesture = features_train()111    print("Test Data","Prediction","True", "Match" )112    # calculate cosine similarity113    results_matrix = np.zeros((17, 17), dtype=int)114    for k in test_gesture:115        test_proba = test_gesture[k].feature_score116        cosine_similarity = float('inf')117        tr_video_file = None118        for key in train_gesture:119            trained = train_gesture[key].feature_score120            cosine_loss = tf.keras.losses.CosineSimilarity(axis=1)121            y = cosine_loss(trained, test_proba).numpy()122            if y < cosine_similarity:123                cosine_similarity = y124                test_gesture[k].predicted_label = train_gesture[key].true_label125                tr_video_file = train_gesture[key].video_name126        results_matrix[test_gesture[k].predicted_label][test_gesture[k].true_label] += 1127        print(test_gesture[k].video_name,test_gesture[k].predicted_label,128            test_gesture[k].true_label, test_gesture[k].is_identified())129    print_result(test_gesture, results_matrix)130if __name__ == '__main__':...

HMM.py

Source:HMM.py

1#By Mohit Minhas2import math3import numpy4#from sklearn.hmm import MultinomialHMM5#from hmmn import *6from hmmpy import *7from sklearn.cluster import k_means8#from scipy.cluster.vq import kmeans29def get_xyz_data(path,name):10    xfl=path+'\\'+name+'_x.csv'11    xx = numpy.genfromtxt(xfl, delimiter=',')12    yfl=path+'\\'+name+'_y.csv'13    xy = numpy.genfromtxt(yfl, delimiter=',')14    zfl=path+'\\'+name+'_z.csv'15    xz = numpy.genfromtxt(zfl, delimiter=',')16    x=[]17    x.append(xx)18    x.append(xy)19    x.append(xz)20    x=numpy.array(x)21    return x22"""23def emprob(M,N):24    a=1/float(N)25    E=numpy.zeros((M,N))26    for i in xrange(M):27        for j in xrange(N):28            E[i][j]=a29    return E30"""31def prior_transition_matrix(K,LR):32    P = numpy.multiply(1/float(LR),numpy.identity(K+1))33    w=1/float(LR)34    for i in xrange(1,K-(LR-1)+1): 35        for j in xrange(1,LR-1+1):36            P[i][i+j]=w37    for i in xrange(K-(LR-2),K+1):38        for j in xrange(1,K-i+1+1):39            P[i][i+(j-1)] = 1/float(K-i+1)40    P=P[1:,1:]41    return P42def get_point_centroids(indata,K,D):43    mean = numpy.zeros((indata.shape[1],D))44    for n in xrange(0,(indata.shape[1])):45        for i in xrange(0,(indata.shape[2])):46            for j in xrange(0,D):47                mean[n][j] = mean[n][j] + indata[j][n][i]48        mean[n] = mean[n]/(indata.shape[2])49    (centroids,x,y)=k_means(mean,K) #random order. change n_jobs to speed up50    return centroids51def get_point_clusters(data,centroids,D):52    XClustered = [[] for x in xrange(data.shape[2])]53    K = centroids.shape[0]54    for n in xrange(0,(data.shape[1])):55        for i in xrange(0,(data.shape[2])):56            temp = numpy.zeros((K,1))57            for j in xrange(0,K):58                #if (D==3)59                temp[j] = math.sqrt(math.pow((centroids[j][0] - data[0][n][i]),2)+math.pow((centroids[j][1] - data[1][n][i]),2)+math.pow((centroids[j][2] - data[2][n][i]),2));60            I = numpy.argmin(temp)61            XClustered[i].append(I)62    XClustered=numpy.array(XClustered)63    return XClustered64def pr_hmm(o,a,b,pi):65    n=len(a[0])66    T=len(o)67    for i in xrange(1,n+1):68        m[1][i]=b[i][o[1]]*pi[i];69    for t in xrange(1,(T-1)+1):70        for j in xrange(1,n+1):71            z=072            for i in xrange(1,n+1):73                z=z+a[i][j]*m[t][i]74            m[t+1][j]=z*b[j][o[t+1]]75    p=076    for i in xrange(1,n+1):77        p=p+m[T][i]        78    p=math.log(p)79    return p80D=381M=1282N=883LR=284train_gesture='x'85test_gesture='x'86gestureRecThreshold = 087training = get_xyz_data('data/train',train_gesture)88testing = get_xyz_data('data/test',test_gesture)89centroids = get_point_centroids(training,N,D)90ATrainBinned = get_point_clusters(training,centroids,D)91ATestBinned = get_point_clusters(testing,centroids,D)92pP = prior_transition_matrix(M,LR)93#W=emprob(M,N)94#print ATrainBinned95#model=MultinomialHMM(n_components=M,startprob_prior=pP,n_iter=50)96#model.n_symbols=N97#print model.n_symbols98#model.fit(ATrainBinned)99#model=MultinomialHMM(ATrainBinned,pP,[1:N]',M,cyc,.00001) #ENTER 100#logprob=model.score(ATestBinned)101#print logprob102hmm=HMM(n_states=M,V=[0,1,2,3,4,5,6,7],A=pP)103print 'TRAINING'104print105baum_welch(hmm,ATrainBinned,graph=False,verbose=True)106print 107print 'TESTING'108print109b=forward(hmm,ATestBinned[0])110print b111#model=DiscreteHmm(numstates=M,numclasses=N)112#model.learn(ATrainBinned,numsteps=ATrainBinned.shape[0])113"""114sumLik = 0.0115minLik = float('inf')116for j in xrange(0,len(ATrainBinned)):117	lik = pr_hmm(ATrainBinned[j],P,E.T,Pi)118	if (lik < minLik):119		minLik = lik120	sumLik = sumLik + lik121	122gestureRecThreshold = 2.0*sumLik/len(ATrainBinned)123print'\n\n********************************************************************\n'124print'Testing {0) sequences for a log likelihood greater than {1)\n'.format(len(ATestBinned),gestureRecThreshold)125print'********************************************************************\n\n'126recs = 0127tLL = numpy.zeros((len(ATestBinned),1))128for j in xrange(1,len(ATestBinned)):129	tLL[j][1] = pr_hmm(ATestBinned[j],P,E.T,Pi)130	if (tLL[j][1] > gestureRecThreshold):131		recs = recs + 1132		print 'Log likelihood: {0) > {1) (threshold) -- FOUND {2) GESTURE!\n'.format(tLL[j][1],gestureRecThreshold,test_gesture)133	else:134		print 'Log likelihood: {0} < {1} (threshold) -- NO {2} GESTURE.\n'.format(tLL[j][1],gestureRecThreshold,test_gesture)135                          136print'Recognition success rate: {0) percent\n'.format(100*recs/len(ATestBinned))...

Automation Testing Tutorials

Learn to execute automation testing from scratch with LambdaTest Learning Hub. Right from setting up the prerequisites to run your first automation test, to following best practices and diving deeper into advanced test scenarios. LambdaTest Learning Hubs compile a list of step-by-step guides to help you be proficient with different test automation frameworks i.e. Selenium, Cypress, TestNG etc.