def ensemble_models(models, id):
_, test_ids = load_test_dataset(size=64)
predictions = np.zeros((test_ids.shape[0], 10))
submission = np.empty((test_ids.shape[0]))
count = 0
for model_data in models:
x_test, _ = load_test_dataset(size=int(model_data[0][-2:]))
print("Predict {}".format(model_data[0]))
count += 1
model = load_model("../{}/models/model_{}.h5".format(
model_data[0], model_data[1]))
prediction = model.predict(x_test)
predictions += prediction
predictions = np.argmax(predictions / count, axis=1)
for label, image_id in zip(predictions, test_ids):
submission[image_id - 1] = label + 1
submission_csv = pd.read_csv(
"/home/shouki/Desktop/Programming/Python/AI/Datasets/ImageData/ISSM/sampleSolution.csv"
)
submission_csv["LABEL"] = np.array(submission, dtype=np.int32)
submission_csv.to_csv("./submission/submission_{}.csv".format(id),
index=False)
def create_submission_file(epoch):
x_test, test_ids = load_test_dataset()
model = load_model("./models/model_{}.h5".format(epoch))
predictions = np.argmax(model.predict(x_test), axis=1)
submission = np.empty((predictions.shape[0]))
for label, image_id in zip(predictions, test_ids):
submission[image_id - 1] = label + 1
submission_csv = pd.read_csv("/home/shouki/Desktop/Programming/Python/AI/Datasets/ImageData/ISSM/sampleSolution.csv")
submission_csv["LABEL"] = np.array(submission, dtype=np.int32)
submission_csv.to_csv("./submission/submission_{}.csv".format(epoch), index=False)
from issm import load_test_dataset
from keras.utils import to_categorical
from keras.models import load_model
import tensorflow as tf
import numpy as np
import pandas as pd
physical_devices = tf.config.experimental.list_physical_devices("GPU")
tf.config.experimental.set_memory_growth(physical_devices[0], True)
x_test, test_ids = load_test_dataset()
model = load_model("./models/model_final.h5")
predictions = np.argmax(model.predict( x_test), axis=1)
submission = np.empty((predictions.shape[0]))
for label, image_id in zip(predictions, test_ids):
submission[image_id - 1] = label + 1
submission_csv = pd.read_csv("/home/shouki/Desktop/Programming/Python/AI/Datasets/ImageData/ISSM/sampleSolution.csv")
submission_csv["LABEL"] = np.array(submission, dtype=np.int32)
submission_csv.to_csv("./submission/submission.csv", index=False)