ByExampleSynthesisOfVectorTextures/optimize_weights.py at main · ChrisPGraphics/ByExampleSynthesisOfVectorTextures · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
import datetime
import logging
import os

import numpy as np
from pymoo.algorithms.soo.nonconvex.ga import GA
from pymoo.core.callback import Callback
from pymoo.core.problem import ElementwiseProblem
from pymoo.optimize import minimize
from pymoo.termination import get_termination

import analysis
import common
import synthesis


class WeightOptimizer(ElementwiseProblem):
    def __init__(
            self, texton_library: analysis.result_objects.PrimaryTextonResult, synthetic_size
    ):
        self.texton_library = texton_library
        self.synthetic_size = synthetic_size
        self.target_global_coverage = self.texton_library.global_coverage

        parameter_count = len(synthesis.Weights().to_array())
        super().__init__(
            n_var=parameter_count,
            xu=np.full(parameter_count, 1),
            xl=np.full(parameter_count, -1),
        )

    def _evaluate(self, x, out, *args, **kwargs):
        weights = synthesis.Weights.from_array(x)

        primary_textons = synthesis.primary_texton_distro(
            self.texton_library.primary_textons, self.synthetic_size, weights=weights,
        )

        actual_distro = synthesis.primary_textons.get_coverage(primary_textons)
        out["F"] = (actual_distro - self.target_global_coverage) ** 2


class WeightOptimizerCallback(Callback):
    def __init__(self, callback=None, patience: int = 0, termination=None, min_error: float = 0) -> None:
        super().__init__()
        self.printed_header = False
        self.start = datetime.datetime.now()
        self.last_score = None
        self.no_improvement = 0
        self.callback = callback
        self.patience = patience
        self.termination = termination
        self.min_error = min_error

    def notify(self, algorithm: GA):
        best_score = algorithm.pop.get("F")[0][0]

        timestamp = str(datetime.datetime.now() - self.start)
        timestamp = timestamp[:timestamp.index(".")]

        if not self.printed_header:
            print(" ELAPSED  GEN       SCORE     CHANGE  PARAMETERS")
            self.printed_header = True

        print("{:>8} {:>4} {:>11.8f}  {:>9.6f}  [{}]".format(
            timestamp, algorithm.n_gen, best_score,
            0 if self.last_score is None else best_score - self.last_score,
            ", ".join("{:>9.6f}".format(i) for i in algorithm.pop.get("X")[0])
        ))

        if self.last_score == best_score:
            self.no_improvement += 1
            if self.no_improvement > self.patience:
                print("No improvement in {} generations. Terminating...".format(self.patience))
                self.termination.terminate()
                self.termination.perc = 1

        else:
            self.no_improvement = 0

        if best_score <= self.min_error:
            print(
                "Found solution with score {:.8f} (below {}). Terminating...".format(best_score, self.min_error)
            )
            self.termination.terminate()
            self.termination.perc = 1

        self.last_score = best_score


def optimize_weights(config: synthesis.OptimizeConfig) -> synthesis.Weights:
    logging.info("Reading analyzed files...")
    extracted_primary_textons = analysis.result_objects.PrimaryTextonResult.load(os.path.join(config.intermediate_directory, "primary_textons.dat"))

    problem = WeightOptimizer(extracted_primary_textons, (config.result_size, config.result_size))

    algorithm = GA(
        pop_size=config.population_size,
        sampling=synthesis.Weights().to_array(),
        eliminate_duplicates=True
    )

    termination = get_termination("n_gen", config.generations)

    start = datetime.datetime.now()
    res = minimize(
        problem,
        algorithm,
        termination,
        save_history=False,
        callback=WeightOptimizerCallback(
            patience=config.patience, termination=termination, min_error=config.min_error
        ),
        copy_termination=False
    )
    end = datetime.datetime.now()

    best_parameters = res.X
    best_score = res.F[0]

    weights = synthesis.Weights.from_array(best_parameters)
    weights.to_json(os.path.join(config.intermediate_directory, "placement_weights.json"), best_score)

    return weights


if __name__ == '__main__':
    common.logger.configure_logger(level=logging.ERROR)
    cfg = synthesis.OptimizeConfig.from_argv()
    optimize_weights(cfg)