data_amount = 1
soft_labels = True
if soft_labels:
from PrepareDataSoftLabels import PrepareData
else:
from PrepareOriginalData import PrepareData
# Load training set
# Note: this operation is needed only to generate a dictionary, but one could also load a saved one instead
p = PrepareData(
path_images='data_vqa_feat', # Path to image features
subset='trainval', # Desired subset: either train2014, val2014 or trainval
taskType='OpenEnded', # 'OpenEnded', 'MultipleChoice', 'all'
cut_data=
data_amount, # Percentage of data to use, 1 = All values, above 1=#samples for debugging
output_path='data', # Path where we want to output temporary data
pad_length=32, # Number of words in a question (zero padded)
question_threshold=6,
answer_threshold=3, # Keep only most common words
questions_sparse=False,
image_extractor='RawImages')
image_features, questions, answers, annotations = p.load_data()
print("Image features", image_features.shape)
print("Question features", questions.shape)
print("Answers", answers.shape)
print("Dictionary size", p.dic_size)
p.dumpDictionary()
# Define neural network
if soft_labels:
neuralnet = NeuralNetwork(image_features.shape[0],
from NeuralNetworkSoftLabels import NeuralNetwork
import numpy as np
from EvaluateModel import ProduceResult
# Some constants
taskType = 'all'
data_amount = 1
epochs = 50
# Load training set
p = PrepareData(
path_images='data_vqa_feat', # Path to image features
subset='trainval', # Desired subset: either train2014 or val2014
taskType='OpenEnded', # 'OpenEnded', 'MultipleChoice', 'all'
cut_data=
data_amount, # Percentage of data to use, 1 = All values, above 1=#samples for debugging
output_path='data', # Path where we want to output temporary data
pad_length=32, # Number of words in a question (zero padded)
question_threshold=0,
answer_threshold=0, # Keep only most common words
questions_sparse=True)
image_features, questions, answers, annotations = p.load_data()
print("Image features", image_features.shape)
print("Question features", questions.shape)
print("Answers", answers.shape)
print("Dictionary size", p.dic_size)
# Save dictionary
p.dumpDictionary()
# Use this when using sparse representation