def run(self):
logging.info("generating population...")
self.population = Population(self.popsize, self.fitness_object, self.min_d, self.max_d)
generations = 0
logging.info("starting...")
# Run first generation without elitism
self.population.next_generation(0, 0, self.recombination_rate, self.mutation_rate, self.selector)
best_value = self.population.generation.individuals[0].get_y()
logging.info("Generation {0}: \n {1}".format(generations, self.population.generation))
generations += 1
# Run next generations with elitism
while generations < self.max_gen and best_value > self.threshold:
self.population.next_generation(self.elites,
self.elite_recombinations,
self.recombination_rate,
self.mutation_rate,
self.selector)
logging.info("Generation {0}: \n{1}".format(generations, self.population.generation))
best_value = self.population.generation.individuals[0].get_y()
generations += 1
for i in range(10):
self.population.next_final_generation(self.recombination_rate, self.selector)
logging.info("Final Generation {0}: \n{1}".format(i, self.population.generation))
logging.info("Best clips: \n{0}".format(self.population.generation))
class LoopingGifFinder(object):
def __init__(self,
clip,
min_d=None,
max_d=None,
threshold=None,
max_gen=None,
popsize=None,
elites=None,
elite_recombinations=None,
recombination_rate=None,
mutation_rate=None):
self.clip = mpy.VideoFileClip(clip).resize(width=20)
self.min_d = 1.5 if min_d is None else min_d
self.max_d = 5 if max_d is None else max_d
self.threshold = 40 if threshold is None else threshold
self.max_gen = 50 if max_gen is None else max_gen
self.popsize = 20 if popsize is None else popsize
self.elites = 2 if elites is None else elites
self.elite_recombinations = 0 if elite_recombinations is None else elite_recombinations
self.recombination_rate = 0.7 if recombination_rate is None else recombination_rate
self.mutation_rate = 0.8 if mutation_rate is None else mutation_rate
self.fitness_object = FrameDistance(self.clip, self.min_d, self.max_d)
self.selector = TournamentSelector(self.popsize)
self.population = None
logging.info("Parameters: \n" +
"\tmin_d: {0}\n".format(self.min_d) +
"\tmax_d: {0}\n".format(self.max_d) +
"\tthreshold: {0}\n".format(self.threshold) +
"\tmax_gen: {0}\n".format(self.max_gen) +
"\tpopsize: {0}\n".format(self.popsize) +
"\telites: {0}\n".format(self.elites) +
"\telite_recombinations: {0}\n".format(self.elite_recombinations) +
"\trecombination_rate: {0}\n".format(self.recombination_rate) +
"\tmutation_rate: {0}".format(self.mutation_rate)
)
def run(self):
logging.info("generating population...")
self.population = Population(self.popsize, self.fitness_object, self.min_d, self.max_d)
generations = 0
logging.info("starting...")
# Run first generation without elitism
self.population.next_generation(0, 0, self.recombination_rate, self.mutation_rate, self.selector)
best_value = self.population.generation.individuals[0].get_y()
logging.info("Generation {0}: \n {1}".format(generations, self.population.generation))
generations += 1
# Run next generations with elitism
while generations < self.max_gen and best_value > self.threshold:
self.population.next_generation(self.elites,
self.elite_recombinations,
self.recombination_rate,
self.mutation_rate,
self.selector)
logging.info("Generation {0}: \n{1}".format(generations, self.population.generation))
best_value = self.population.generation.individuals[0].get_y()
generations += 1
for i in range(10):
self.population.next_final_generation(self.recombination_rate, self.selector)
logging.info("Final Generation {0}: \n{1}".format(i, self.population.generation))
logging.info("Best clips: \n{0}".format(self.population.generation))
def results_to_gif(self, prefix):
for i in range(10):
self.clip.subclip(*self.population.generation.individuals[i].x).write_gif("{0}_{1}.gif".format(prefix, i))
