Questo script prende in input il file “shuffled_output_data.csv”, che è il file input della fase di evaluation, e restituisce un file che contiene come nome dell’entità le sue lexical references. Il file quindi risulta essere molto più massiccio di “shuffled_output_data.csv”. Verrà, infatti, troncato alle prime 10k righe.
| environment | entity_type | image_bbox | image_normal | bounding_box |
|---|---|---|---|---|
| 2380 | phone | Robocup/2380/images/LivingRoom/bounding_box/position_1/2380_LivingRoom_bounding_box_pos_1_180.jpg | Robocup/2380/images/LivingRoom/normal/position_1/2380_LivingRoom_pos_1_180.jpg | (0.465, 0.31333333333333335, 0.475, 0.31333333333333335) |
| 2304 | keys | Robocup/2304/images/Bedroom/bounding_box/position_1/2304_Bedroom_bounding_box_pos_1_90.jpg | Robocup/2304/images/Bedroom/normal/position_1/2304_Bedroom_pos_1_90.jpg | (0.0, 0.425, 0.006666666666666667, 0.43) |
| 3489 | jar | Release1/3489/images/Kitchen/bounding_box/position_6/3489_Kitchen_bounding_box_pos_6_0.jpg | Release1/3489/images/Kitchen/normal/position_6/3489_Kitchen_pos_6_0.jpg | (0.5583333333333333, 0.2683333333333333, 0.575, 0.29) |
| 2192 | jam | Robocup/2192/images/Kitchen/bounding_box/position_1/2192_Kitchen_bounding_box_pos_1_90.jpg | Robocup/2192/images/Kitchen/normal/position_1/2192_Kitchen_pos_1_90.jpg | (0.49166666666666664, 0.375, 0.52, 0.39) |
import pandas as pd
# Read the shuffled output data CSV file
df = pd.read_csv("shuffled_output_data.csv")
# Initialize an empty list to store rows
new_rows = []
# Iterate through each row of the DataFrame
for index, row in df.iterrows():
# Split the lexical_references string into a list
lexical_references = eval(row['lexical_references'])
# If there are lexical references, create a new row for each one
if lexical_references:
for reference in lexical_references:
# Create a new row with the lexical reference as the entity type
new_row = row.copy()
new_row['entity_type'] = reference
# Append the new row to the list of new rows
new_rows.append(new_row)
else:
# If there are no lexical references, keep the original row
new_rows.append(row)
# Create a new DataFrame from the list of new rows
new_df = pd.DataFrame(new_rows)
# Drop the 'lexical_references' column
new_df.drop(columns=['lexical_references'], inplace=True)
# Write the new DataFrame to a new CSV file
new_df.to_csv("expanded_output_data.csv", index=False)