def load_data(dataset): if dataset == 'test': X, y = load_test() sz = 230 elif dataset == 'uc1': X, y = split_df(pd.read_pickle('..\\data\\df_uc1.pkl'), index_column='run_id', feature_columns=['fldPosition', 'fldCurrent'], target_name='target') # Length of timeseries for resampler and cnn sz = 38 elif dataset == 'uc2': X, y = split_df(pd.read_pickle('..\\data\\df_uc2.pkl'), index_column='run_id', feature_columns=['position', 'force'], target_name='label') # Length of timeseries for resampler and cnn sz = 200 resampler = TimeSeriesResampler(sz=sz) X = resampler.fit_transform(X, y) y = np.array(y) return X, y
from alpaca import Alpaca from utils import to_time_series_dataset, to_dataset, split_df, TimeSeriesResampler import time import numpy as np import pandas as pd from sklearn.pipeline import Pipeline max_sample = 20 for dataset in ['uc2']: if dataset == 'uc1': X, y = split_df(pd.read_pickle('..\\data\\df_uc1.pkl'), index_column='run_id', feature_columns=['fldPosition', 'fldCurrent'], target_name='target') y = np.array(y) # Length of timeseries for resampler and cnn sz = 38 # Number of channels for cnn num_channels = len(X[0][0]) # Number of classes for cnn num_classes = np.unique(y).shape[0] if dataset == 'uc2': X, y = split_df(pd.read_pickle('..\\data\\df_uc2.pkl'), index_column='run_id', feature_columns=['position', 'force'], target_name='label') y = np.array(y) # Length of timeseries for resampler and cnn sz = 200 # Number of channels for cnn
import utils df_pickled_path = "./data.pkl" data = utils.load_df_pickled(df_pickled_path) train_data, valid_data, test_data = utils.split_df(data, train_set_ratio=0.8, test_set_ratio=0.1, valid_set_ratio=0.1) if __name__ == "__main__": pass
print(f'The list different size image: {heigh_width_unique}') # Train file annot_df = pd.DataFrame.from_records(train_json['annotations']) annot_df['weight'] = heigh_width_unique[0][1] annot_df['height'] = heigh_width_unique[0][0] annot_df['x'] = annot_df['bbox'].apply(lambda x: x[0]) annot_df['y'] = annot_df['bbox'].apply(lambda x: x[1]) annot_df['w'] = annot_df['bbox'].apply(lambda x: x[2]) annot_df['h'] = annot_df['bbox'].apply(lambda x: x[3]) display(annot_df.head()) # List label categorie_df = pd.DataFrame.from_records(train_json['categories']) display((categorie_df)) # List image in folder train_image_path = glob(f'{args.train_image}/*.*') test_image_path = glob(f'{args.test_image}/*.*') print( f'Number of train image: {len(train_image_path)}, test image: {len(test_image_path)}' ) # Create fold annot_pivot = split_df(annot_df) display(annot_pivot.head()) #Train process train_pr = Train_process() train_pr.fit(annot_df, annot_pivot)