Commit d27fe6fcc52cc5c3a4ea61289d4d4b0c38c53b83
1 parent
acbafc4147
Exists in
master
add utt2dur statistics command and changed original stats command by a subcommand pred-distribution
Showing 1 changed file with 55 additions and 23 deletions Inline Diff
volia/stats.py
| 1 | 1 | ||
| 2 | import argparse | 2 | import argparse |
| 3 | 3 | ||
| 4 | import os | 4 | import os |
| 5 | import core.data | 5 | import core.data |
| 6 | import math | 6 | import math |
| 7 | import numpy as np | 7 | import numpy as np |
| 8 | import scipy.stats | 8 | import scipy.stats |
| 9 | import pickle | 9 | import pickle |
| 10 | import matplotlib.pyplot as plt | 10 | import matplotlib.pyplot as plt |
| 11 | import matplotlib.colors as mcolors | 11 | import matplotlib.colors as mcolors |
| 12 | from utils import SubCommandRunner | ||
| 12 | 13 | ||
| 13 | 14 | ||
| 14 | |||
| 15 | from cycler import cycler | 15 | from cycler import cycler |
| 16 | 16 | ||
| 17 | def stats(): | 17 | def pred_distribution(predictions: str, labels: str, labelencoder: str, outdir: str): |
| 18 | print("Decisions") | ||
| 19 | 18 | ||
| 20 | |||
| 21 | print(list(mcolors.TABLEAU_COLORS)) | ||
| 22 | |||
| 23 | |||
| 24 | if __name__ == "__main__": | ||
| 25 | |||
| 26 | # Parser | ||
| 27 | parser = argparse.ArgumentParser(description="") | ||
| 28 | |||
| 29 | # Arguments | ||
| 30 | parser.add_argument("--predictions", type=str, help="prediction file", required=True) | ||
| 31 | parser.add_argument("--labels", type=str, help="label file", required=True) | ||
| 32 | parser.add_argument("--labelencoder", type=str, help="label encode pickle file", required=True) | ||
| 33 | parser.add_argument("--outdir", type=str, help="output file", required=True) | ||
| 34 | |||
| 35 | args = parser.parse_args() | ||
| 36 | |||
| 37 | predictions = core.data.read_id_values(args.predictions, float) | 19 | predictions = core.data.read_id_values(args.predictions, float) |
| 38 | labels = core.data.read_labels(args.labels) | 20 | labels = core.data.read_labels(args.labels) |
| 39 | 21 | ||
| 40 | le = None | 22 | le = None |
| 41 | with open(args.labelencoder, "rb") as f: | 23 | with open(args.labelencoder, "rb") as f: |
| 42 | le = pickle.load(f) | 24 | le = pickle.load(f) |
| 43 | stats = {} | 25 | stats = {} |
| 44 | 26 | ||
| 45 | print("PREDICTIONS ---------------------------") | 27 | print("PREDICTIONS ---------------------------") |
| 46 | for id_, predictions_ in predictions.items(): | 28 | for id_, predictions_ in predictions.items(): |
| 47 | label = labels[id_][0] | 29 | label = labels[id_][0] |
| 48 | if label not in stats: | 30 | if label not in stats: |
| 49 | stats[label] = { | 31 | stats[label] = { |
| 50 | "nb_utt": 1, | 32 | "nb_utt": 1, |
| 51 | "predictions": np.expand_dims(predictions_, axis=0) | 33 | "predictions": np.expand_dims(predictions_, axis=0) |
| 52 | } | 34 | } |
| 53 | else: | 35 | else: |
| 54 | stats[label]["nb_utt"] = stats[label]["nb_utt"] + 1 | 36 | stats[label]["nb_utt"] = stats[label]["nb_utt"] + 1 |
| 55 | stats[label]["predictions"] = np.append(stats[label]["predictions"], np.expand_dims(predictions_, axis=0), axis=0) | 37 | stats[label]["predictions"] = np.append(stats[label]["predictions"], np.expand_dims(predictions_, axis=0), axis=0) |
| 56 | 38 | ||
| 57 | 39 | ||
| 58 | print("CALCULATING ---------------------------") | 40 | print("CALCULATING ---------------------------") |
| 59 | 41 | ||
| 60 | 42 | ||
| 61 | colors = [ | 43 | colors = [ |
| 62 | "darkorange", | 44 | "darkorange", |
| 63 | "red", | 45 | "red", |
| 64 | "blue" | 46 | "blue" |
| 65 | ] | 47 | ] |
| 66 | custom_cycler = (cycler(color=list(mcolors.TABLEAU_COLORS)) * | 48 | custom_cycler = (cycler(color=list(mcolors.TABLEAU_COLORS)) * |
| 67 | cycler(linestyle=['-', '--', '-.'])) | 49 | cycler(linestyle=['-', '--', '-.'])) |
| 68 | 50 | ||
| 69 | 51 | ||
| 70 | for label, stats_ in stats.items(): | 52 | for label, stats_ in stats.items(): |
| 71 | 53 | ||
| 72 | plt.gca().set_prop_cycle(custom_cycler) | 54 | plt.gca().set_prop_cycle(custom_cycler) |
| 73 | stats_mean = np.mean(stats_["predictions"], axis=0) | 55 | stats_mean = np.mean(stats_["predictions"], axis=0) |
| 74 | stats_std = np.std(stats_["predictions"], axis=0) | 56 | stats_std = np.std(stats_["predictions"], axis=0) |
| 75 | 57 | ||
| 76 | #print(label) | 58 | #print(label) |
| 77 | #print(stats_mean) | 59 | #print(stats_mean) |
| 78 | #print(stats_std) | 60 | #print(stats_std) |
| 79 | kwargs = dict(alpha=0.5) | 61 | kwargs = dict(alpha=0.5) |
| 80 | 62 | ||
| 81 | for i in range(stats_["predictions"].shape[1]): | 63 | for i in range(stats_["predictions"].shape[1]): |
| 82 | label_str = le.inverse_transform([i])[0] | 64 | label_str = le.inverse_transform([i])[0] |
| 83 | #plt.hist(stats_["predictions"][:, i], bins=10, label=label_str, **kwargs) | 65 | #plt.hist(stats_["predictions"][:, i], bins=10, label=label_str, **kwargs) |
| 84 | mu = stats_mean[i] | 66 | mu = stats_mean[i] |
| 85 | variance = stats_std[i] * stats_std[i] | 67 | variance = stats_std[i] * stats_std[i] |
| 86 | sigma = stats_std[i] | 68 | sigma = stats_std[i] |
| 87 | # math.sqrt(variance) | 69 | # math.sqrt(variance) |
| 88 | print(f"{i}: mu {mu}, var {variance}, sigma {sigma}") | 70 | print(f"{i}: mu {mu}, var {variance}, sigma {sigma}") |
| 89 | 71 | ||
| 90 | #x_values = np.arange(-1, 5, 0.1) | 72 | #x_values = np.arange(-1, 5, 0.1) |
| 91 | 73 | ||
| 92 | #y_values = scipy.stats.norm(mu, variance) | 74 | #y_values = scipy.stats.norm(mu, variance) |
| 93 | #y = scipy.stats.norm.pdf(x,mean,std) | 75 | #y = scipy.stats.norm.pdf(x,mean,std) |
| 94 | 76 | ||
| 95 | #plt.plot(x_values, y_values.pdf(x_values,)) | 77 | #plt.plot(x_values, y_values.pdf(x_values,)) |
| 96 | 78 | ||
| 97 | #x, step = np.linspace(mu - 3*sigma, mu + 3*sigma, 1000, retstep=True) | 79 | #x, step = np.linspace(mu - 3*sigma, mu + 3*sigma, 1000, retstep=True) |
| 98 | x = np.linspace(0, 1, 1000) | 80 | x = np.linspace(0, 1, 1000) |
| 99 | #x = np.linspace(mu - 3*sigma, mu + 3*sigma, 1000) | 81 | #x = np.linspace(mu - 3*sigma, mu + 3*sigma, 1000) |
| 100 | #x, step = np.linspace(0, 1, 1000, retstep=True) | 82 | #x, step = np.linspace(0, 1, 1000, retstep=True) |
| 101 | 83 | ||
| 102 | P = scipy.stats.norm.cdf(x, mu, sigma) | 84 | P = scipy.stats.norm.cdf(x, mu, sigma) |
| 103 | #print(step) | 85 | #print(step) |
| 104 | plt.plot(x, P, label=label_str, **kwargs) | 86 | plt.plot(x, P, label=label_str, **kwargs) |
| 105 | #plt.savefig("simple_gaussian.pdf") | 87 | #plt.savefig("simple_gaussian.pdf") |
| 106 | 88 | ||
| 107 | plt.legend() | 89 | plt.legend() |
| 108 | plt.savefig(os.path.join(args.outdir, f"{label}_prediction_cdf.pdf")) | 90 | plt.savefig(os.path.join(args.outdir, f"{label}_prediction_cdf.pdf")) |
| 109 | plt.clf() | 91 | plt.clf() |
| 110 | 92 | ||
| 93 | print("Decisions") | ||
| 111 | 94 | ||
| 112 | # TODO: | 95 | |
| 113 | # One graph for each label. Distribution of their predictions output are displayed. | 96 | def utt2dur(utt2dur: str, labels: str): |
| 114 | 97 | if labels == None: | |
| 98 | pass | ||
| 99 | else: | ||
| 100 | pass | ||
| 101 | |||
| 102 | durations = [] | ||
| 103 | with open(utt2dur, "r") as f: | ||
| 104 | for line in f: | ||
| 105 | splited = line.replace("\n", "").split(" ") | ||
| 106 | durations.append(float(splited[1])) | ||
| 107 | |||
| 108 | durations = np.asarray(durations, dtype=float) | ||
| 109 | print(durations.shape) | ||
| 110 | mean = np.mean(durations) | ||
| 111 | std = np.std(durations) | ||
| 112 | |||
| 113 | print(f"mean: {mean}") | ||
| 114 | print(f"std: {std}") | ||
| 115 | |||
| 116 | |||
| 117 | if __name__ == "__main__": | ||
| 118 | |||
| 119 | # Parser | ||
| 120 | parser = argparse.ArgumentParser(description="Statistics") | ||
| 121 | subparsers = parser.add_subparsers(title="actions") | ||
| 122 | |||
| 123 | # pred-distribution | ||
| 124 | parser_pred_dist = subparsers.add_parser("pred-distribution", help="plot distributions of prediction through labels") | ||
| 125 | parser_pred_dist.add_argument("--predictions", type=str, help="prediction file", required=True) | ||
| 126 | parser_pred_dist.add_argument("--labels", type=str, help="label file", required=True) | ||
| 127 | parser_pred_dist.add_argument("--labelencoder", type=str, help="label encode pickle file", required=True) |