def GA_Version_1(version_path, Runs, generations, parent_pop_init, mating_pop, mutationRate, mutationFunc, passes, instance): minFitnessValsOfAllRuns = [] averageFitnessValsOfAllRuns = [] for r in range(0, Runs): img_path = version_path + "GA instance " + str(instance) + "/Images/" initial_population_path = version_path + "GA instance " + str( instance) + "/initial_population" + " for Run " + str(r + 1) + ".json" final_population_path = version_path + "GA instance " + str( instance) + "/final_population" + " for Run " + str(r + 1) + ".json" data_path = version_path + "GA instance " + str(instance) + "/" random_seed = 100 * (r + 1) np.random.seed(random_seed) LB = 0 UB = 1 initial_image = cv2.imread("binaryhand.png", 0) initialize_population(random_seed, initial_image, parent_pop_init, bonus=False) with open("CellularAutomata.json", 'r') as f: population = json.loads(f.read()) start_image = np.asarray(population[0]) final_image = np.asarray(population[0]) goal_image = np.asarray(population[1]) save_image(img_path, initial_image, "initial_image", r) save_image(img_path, start_image, "start_image", r) save_image(img_path, goal_image, "goal_image", r) goal_image_binary = grayToBinary(goal_image) save_image(img_path, goal_image_binary, "goal_image_binary", r) population_used = copy.deepcopy(population[2:]) save_data(initial_population_path, population_used) s = 0 minFitnessValsPerRun = [] averageFitnessValsPerRun = [] diversity_initial = calculateDiversity(population_used) #print("Initial diversity= "+str(diversity_initial)+" for Run: "+str(r+1)) for epoch in range(0, generations): parent_pop_size = len(population_used) parentPopfitnessVals = evaluatePopulationFitness( population_used, parent_pop_size, passes, final_image, goal_image_binary) [minfitness, bestSolution] = solutionTaggerSorter(population_used, parentPopfitnessVals, parent_pop_size, 1, False) minfitness = minfitness[0] averagefitness = stat.mean(parentPopfitnessVals) matingPool = population_used muArgs = { "matingPool": matingPool, "mutationRate": mutationRate, "LB": LB, "UB": UB } mutationStrategy, args = mutationStrategySelector( mutationFunc, muArgs) lamb = len(matingPool) offsprings = mutationFunction(mutationStrategy, args) offspringFitnessVals = evaluatePopulationFitness( offsprings, lamb, passes, final_image, goal_image_binary) newGenPop = crowding(population_used, offsprings, parentPopfitnessVals, offspringFitnessVals, parent_pop_size) if epoch == 0: initFitness = minfitness population_used = newGenPop averageFitnessValsPerRun.append(averagefitness) minFitnessValsPerRun.append(minfitness) diversity_current = calculateDiversity(newGenPop) # print("diversity of current generation= "+str(diversity_current)+" for Run: "+str(r+1)) # print("epoch="+str(epoch)+" for Run: "+str(r+1)) # print("minfitness="+str(minfitness)+" for Run: "+str(r+1)) # print("\n") # print("initminfitness="+str(initFitness)+" for Run: "+str(r+1)) # print("finalminfitness="+str(minfitness)+" for Run: "+str(r+1)) # print("\n") print(" Run: " + str(r + 1)) save_values(data_path + "data" + " for Run " + str(r + 1) + ".json", [ diversity_initial, diversity_current, initFitness, minfitness, bestSolution ], [ "Initial diversity", "final diversity", "Initial minimum fitness", "final minimum fitness", "Best Solution" ]) save_image(img_path, getFinalImage(start_image, bestSolution, passes), "final_image_binary", r) save_data(final_population_path, population_used) minFitnessValsOfAllRuns.append(minFitnessValsPerRun) averageFitnessValsOfAllRuns.append(averageFitnessValsPerRun) stats = progressOfEvolution(Runs, generations, minFitnessValsOfAllRuns, averageFitnessValsOfAllRuns) return stats
def GA_Version_1(Runs, generations, parent_pop_init, mating_pop, mutationRate, crossoverRate, mutationFunc, crossoverFunc, passes): print("GA version 1 begins here") minFitnessValsOfAllRuns = [] averageFitnessValsOfAllRuns = [] for r in range(0, Runs): random_seed = 100 * (r + 1) np.random.seed(random_seed) LB = 0 UB = 1 initial_image = cv2.imread("binaryhand.png", 0) initialize_population(initial_image, parent_pop_init, bonus=False) with open("CellularAutomata.json", 'r') as f: population = json.loads(f.read()) start_image = np.asarray(population[0]) final_image = np.asarray(population[0]) goal_image = np.asarray(population[1]) visualize_image(initial_image) visualize_image(start_image) visualize_image(goal_image) goal_image_binary = grayToBinary(goal_image) population_used = copy.deepcopy(population[2:]) s = 0 minFitnessValsPerRun = [] averageFitnessValsPerRun = [] diversity_initial = calculateDiversity(population_used) print("Initial diversity= " + str(diversity_initial) + " for Run: " + str(r + 1)) for epoch in range(0, generations): parent_pop_size = len(population_used) parentPopfitnessVals = evaluatePopulationFitness( population_used, parent_pop_size, passes, final_image, goal_image_binary) [minfitness, bestSolution] = solutionTaggerSorter(population_used, parentPopfitnessVals, parent_pop_size, 1, False) minfitness = minfitness[0] averagefitness = stat.mean(parentPopfitnessVals) matingPool = fitnessProportionateSelectionWithoutScaling( population_used, parentPopfitnessVals, s, parent_pop_size, mating_pop) muArgs = { "matingPool": matingPool, "mutationRate": mutationRate, "LB": LB, "UB": UB } crArgs = { "matingPool": matingPool, "mating_pop": mating_pop, "LB": LB, "UB": UB, "crossoverRate": crossoverRate } mutationStrategy, args = mutationStrategySelector( mutationFunc, muArgs) crossoverStrategy, args1 = crossoverStrategySelector( crossoverFunc, crArgs) crossoverFunction(crossoverStrategy, args1) lamb = len(matingPool) offsprings = mutationFunction(mutationStrategy, args) offspringFitnessVals = evaluatePopulationFitness( offsprings, lamb, passes, final_image, goal_image_binary) newGenPop = crowding(population_used, offsprings, parentPopfitnessVals, offspringFitnessVals, parent_pop_size) if epoch == 0: initFitness = minfitness population_used = newGenPop averageFitnessValsPerRun.append(averagefitness) minFitnessValsPerRun.append(minfitness) diversity_current = calculateDiversity(newGenPop) print("diversity of current generation= " + str(diversity_current) + " for Run: " + str(r + 1)) print("epoch=" + str(epoch) + " for Run: " + str(r + 1)) print("minfitness=" + str(minfitness) + " for Run: " + str(r + 1)) print("initminfitness=" + str(initFitness) + " for Run: " + str(r + 1)) print("finalminfitness=" + str(minfitness) + " for Run: " + str(r + 1)) visualize_image(getFinalImage(start_image, bestSolution, passes)) minFitnessValsOfAllRuns.append(minFitnessValsPerRun) averageFitnessValsOfAllRuns.append(averageFitnessValsPerRun) stats = progressOfEvolution(Runs, generations, minFitnessValsOfAllRuns, averageFitnessValsOfAllRuns) print("GA version 1 ends here") return stats
imgs = ['butterfly.png', 'flower.png', 'gates.png', 'simpleshapes.png'] start_imgs = [] goal_imgs = [] img_path = "Test Image Results/" with open("BestRuleTable.json", 'r') as f: ruleTable = json.loads(f.read()) psnr_GA = {} msssim_GA = {} fitness_GA = {} for i in range(4): initial_image = cv2.imread("images resized/" + imgs[i], 0) initialize_population(initial_image, 1, bonus=True) with open("CellularAutomata.json", 'r') as f: population = json.loads(f.read()) start_image = np.asarray(population[0]) goal_image = np.asarray(population[1]) start_image_binary = grayToBinary(start_image) plt.imshow(start_image_binary, cmap="gray") plt.show() goal_image_binary = grayToBinary(goal_image) name = imgs[i][:-4] final_image = getFinalImage(start_image, ruleTable, 1)