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

How to use multi_scale_search method in Airtest

Best Python code snippet using Airtest

play.py

Source:play.py

...8import shutil9import time10import cv211import numpy as np12def multi_scale_search(pivot, screen, range=0.3, num=10):13    H, W = screen.shape[:2]14    h, w = pivot.shape[:2]15    found = None16    for scale in np.linspace(1 - range, 1 + range, num)[::-1]:17        resized = cv2.resize(screen, (int(W * scale), int(H * scale)))18        r = W / float(resized.shape[1])19        if resized.shape[0] < h or resized.shape[1] < w:20            break21        res = cv2.matchTemplate(resized, pivot, cv2.TM_CCOEFF_NORMED)22        loc = np.where(res >= res.max())23        pos_h, pos_w = list(zip(*loc))[0]24        if found is None or res.max() > found[-1]:25            found = (pos_h, pos_w, r, res.max())26    if found is None: return (0, 0, 0, 0, 0)27    pos_h, pos_w, r, score = found28    start_h, start_w = int(pos_h * r), int(pos_w * r)29    end_h, end_w = int((pos_h + h) * r), int((pos_w + w) * r)30    return [start_h, start_w, end_h, end_w, score]31class WechatAutoJump(object):32    def __init__(self, sensitivity, debug, resource_dir):33        self.sensitivity = sensitivity34        self.debug = debug35        self.resource_dir = resource_dir36        self.bb_size = [300, 300]37        self.step = 138        self.load_resource()39        if self.debug:40            if not os.path.exists(self.debug):41                os.mkdir(self.debug)42    def load_resource(self):43        self.player = cv2.imread(os.path.join(self.resource_dir, 'player.png'), 0)44        circle_file = glob.glob(os.path.join(self.resource_dir, 'circle/*.png'))45        table_file = glob.glob(os.path.join(self.resource_dir, 'table/*.png'))46        self.jump_file = [cv2.imread(name, 0) for name in circle_file + table_file]47    def get_current_state(self):48        pic_filename = 'state{:03d}.png'.format(self.step)49        state = cv2.imread(pic_filename)50        self.resolution = state.shape[:2]51        scale = state.shape[1] / 720.52        state = cv2.resize(state, (720, int(state.shape[0] / scale)), interpolation=cv2.INTER_NEAREST)53        if state.shape[0] > 1280:54            s = (state.shape[0] - 1280) // 255            state = state[s:(s + 1280), :, :]56        elif state.shape[0] < 1280:57            s1 = (1280 - state.shape[0]) // 258            s2 = (1280 - state.shape[0]) - s159            pad1 = 255 * np.ones((s1, 720, 3), dtype=np.uint8)60            pad2 = 255 * np.ones((s2, 720, 3), dtype=np.uint8)61            state = np.concatenate((pad1, state, pad2), 0)62        return state63    def get_player_position(self, state):64        state = cv2.cvtColor(state, cv2.COLOR_BGR2GRAY)65        pos = multi_scale_search(self.player, state, 0.3, 10)66        h, w = int((pos[0] + 13 * pos[2]) / 14.), (pos[1] + pos[3]) // 267        return np.array([h, w])68    def get_target_position(self, state, player_pos):69        state = cv2.cvtColor(state, cv2.COLOR_BGR2GRAY)70        sym_center = [1280, 720] - player_pos71        sym_tl = np.maximum([0, 0], sym_center + np.array([-self.bb_size[0] // 2, -self.bb_size[1] // 2]))72        sym_br = np.array(73            [min(sym_center[0] + self.bb_size[0] // 2, player_pos[0]), min(sym_center[0] + self.bb_size[1] // 2, 720)])74        state_cut = state[sym_tl[0]:sym_br[0], sym_tl[1]:sym_br[1]]75        target_pos = None76        for target in self.jump_file:77            pos = multi_scale_search(target, state_cut, 0.4, 15)78            if target_pos is None or pos[-1] > target_pos[-1]:79                target_pos = pos80        return np.array([(target_pos[0] + target_pos[2]) // 2, (target_pos[1] + target_pos[3]) // 2]) + sym_tl81    def get_target_position_fast(self, state, player_pos):82        state_cut = state[:player_pos[0], :, :]83        m1 = (state_cut[:, :, 0] == 245)84        m2 = (state_cut[:, :, 1] == 245)85        m3 = (state_cut[:, :, 2] == 245)86        m = np.uint8(np.float32(m1 * m2 * m3) * 255)87        b1, b2 = cv2.connectedComponents(m)88        for i in range(1, np.max(b2) + 1):89            x, y = np.where(b2 == i)90            # print('fast', len(x))91            if len(x) > 280 and len(x) < 310:...

main.py

Source:main.py

1import cv2;2import numpy as np3import os, glob, shutil4import random5def multi_scale_search(pivot, screen, range=0.3, num=10):6    H, W = screen.shape[:2]7    h, w = pivot.shape[:2]8    found = None9    for scale in np.linspace(1-range, 1+range, num)[::-1]:10        resized = cv2.resize(screen, (int(W * scale), int(H * scale)))11        r = W / float(resized.shape[1])12        if resized.shape[0] < h or resized.shape[1] < w:13            break14        res = cv2.matchTemplate(resized, pivot, cv2.TM_CCOEFF_NORMED)15        loc = np.where(res >= res.max())16        pos_h, pos_w = list(zip(*loc))[0]17        if found is None or res.max() > found[-1]:18            found = (pos_h, pos_w, r, res.max())19    if found is None: return (0,0,0,0,0)20    pos_h, pos_w, r, score = found21    start_h, start_w = int(pos_h * r), int(pos_w * r)22    end_h, end_w = int((pos_h + h) * r), int((pos_w + w) * r)23    return [start_h, start_w, end_h, end_w, score]24class wechat_jump(object):25	def __init__(self):26		self.resource_dir = "./resources"27		self.sensitivity = 2.04528		self.bb_size = [300, 300]29		self.load_resources()30		31	def load_resources(self):32		self.player = cv2.imread(os.path.join(self.resource_dir + '/position/player.png'), 0)33		circle_file = glob.glob(os.path.join(self.resource_dir + '/position/circle/*.png'))34		table_file  = glob.glob(os.path.join(self.resource_dir + '/position/table/*.png'))35		self.jump_file = [cv2.imread(name, 0) for name in circle_file + table_file]36	def get_player_position(self, state):37		state = cv2.cvtColor(state, cv2.COLOR_BGR2GRAY)38		pos = multi_scale_search(self.player, state, 0.3, 10)39		h, w = int((pos[0] + 13 * pos[2])/14.), (pos[1] + pos[3])//240		return np.array([h, w])41	def get_target_position_fast(self, state, player_pos):42		state_cut = state[:player_pos[0],:,:]43		m1 = (state_cut[:, :, 0] == 245)44		m2 = (state_cut[:, :, 1] == 245)45		m3 = (state_cut[:, :, 2] == 245)46		m = np.uint8(np.float32(m1 * m2 * m3) * 255)47		b1, b2 = cv2.connectedComponents(m)48		for i in range(1, np.max(b2) + 1):49			x, y = np.where(b2 == i)50			# print('fast', len(x))51			if len(x) > 280 and len(x) < 310:52				r_x = []53				r_y = x, y54		h, w = int(r_x.mean()), int(r_y.mean())55		return np.array([h, w])56	def get_target_position(self, state, player_pos):57		state = cv2.cvtColor(state, cv2.COLOR_BGR2GRAY)58		sym_center = [1280, 720] - player_pos59		sym_tl = np.maximum([0,0], sym_center + np.array([-self.bb_size[0]//2, -self.bb_size[1]//2]))60		sym_br = np.array([min(sym_center[0] + self.bb_size[0]//2, player_pos[0]), min(sym_center[0] + self.bb_size[1]//2, 720)])61		state_cut = state[sym_tl[0]:sym_br[0], sym_tl[1]:sym_br[1]]62		target_pos = None63		for target in self.jump_file:64			pos = multi_scale_search(target, state_cut, 0.4, 15)65			if target_pos is None or pos[-1] > target_pos[-1]:66				target_pos = pos67		return np.array([(target_pos[0]+target_pos[2])//2, (target_pos[1]+target_pos[3])//2]) + sym_tl68	def get_state(self):69		# state image70		# os.system('adb shell screencap -p /sdcard/state.png')71		# os.system('adb pull /sdcard/state.png ' + self.resource_dir + '/screen/state.png')72		state = cv2.imread(self.resource_dir + '/screen/state.png')73		self.resolution = state.shape[:2]74		scale = state.shape[1] / 720.075		state = cv2.resize(state, (720, int(state.shape[0] / scale)), interpolation=cv2.INTER_NEAREST)76		if state.shape[0] > 1280:77			s = state.shape[0] - 128078			state = state[s:,:,:]...

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.