class FvqaTestDataset(Dataset):
def __init__(self, config, overfit=False):
super().__init__()
self.que_vocabulary = Vocabulary(config["dataset"]["word2id_path"])
self.image_features = []
self.image_boxes = []
self.image_labels = []
self.image_captions = []
self.questions = []
self.questions_ids = []
self.questions_vecs = []
self.answers = []
self.config = config
with open(config['dataset']['test']['test_qids'], 'r') as f:
self.test_qids = json.load(f)
with open(config['dataset']['test']['test_questions'], 'r') as f:
self.test_questions = json.load(f)
with open(config['dataset']['test']['test_answers'], 'r') as f:
self.test_answers = json.load(f)
with open(config['dataset']['test']['test_gt_facts'], 'r') as f:
self.test_gt_facts = json.load(f)
with open(config['dataset']['test']['test_captions'], 'r') as f:
self.test_captions = json.load(f)
with open(config['dataset']['test']['test_semantic_graph'], 'rb') as f:
self.semantic_graphs = pickle.load(f, encoding='iso-8859-1')
with open(config['dataset']['test']['test_labels'], 'r') as f:
self.test_labels = json.load(f)
with open(config['dataset']['test']['test_whs'], 'r') as f:
self.test_whs = json.load(f)
with open(config['dataset']['test']['test_top100_facts_graph'],
'rb') as f:
self.test_top_facts = pickle.load(f, encoding='iso-8859-1')
self.test_features = np.load(
config['dataset']['test']['test_features'])
self.test_bboxes = np.load(config['dataset']['test']['test_bboxes'])
if overfit:
self.test_qids = self.test_qids[:100]
self.test_questions = self.test_questions[:100]
self.test_answers = self.test_answers[:100]
self.test_gt_facts = self.test_gt_facts[:100]
self.test_top_facts = self.test_top_facts[:100]
self.test_captions = self.test_captions[:100]
self.test_bboxes = self.test_bboxes[:100]
self.test_features = self.test_features[:100]
self.test_whs = self.test_whs[:100]
self.semantic_graphs = self.semantic_graphs[:100]
def __getitem__(self, index):
test_id = self.test_qids[index]
test_question = self.test_questions[index]
test_answer = self.test_answers[index]
test_gt_fact = self.test_gt_facts[index]
test_top_facts = self.test_top_facts[index]
test_captions = self.test_captions[index]
test_bboxes = self.test_bboxes[index] # (36,4)
test_features = torch.tensor(self.test_features[index])
semantic_graph = self.semantic_graphs[index]
w = self.test_whs[index][0]
h = self.test_whs[index][1]
img_relations = torch.zeros(36, 36, 7)
for i in range(36):
for j in range(36):
xi = test_bboxes[i][0]
yi = test_bboxes[i][1]
wi = test_bboxes[i][2]
hi = test_bboxes[i][3]
xj = test_bboxes[j][0]
yj = test_bboxes[j][1]
wj = test_bboxes[j][2]
hj = test_bboxes[j][3]
r1 = (xj - xi) / (wi * hi)**0.5
r2 = (yj - yi) / (wi * hi)**0.5
r3 = wj / wi
r4 = hj / hi
r5 = (wj * hj) / wi * hi
r6 = sqrt((xj - xi)**2 + (yj - yi)**2) / sqrt(w**2 + h**2)
r7 = atan2(yj - yi, xj - xi)
rel = torch.tensor([r1, r2, r3, r4, r5, r6, r7])
img_relations[i][j] = rel
# 归一化
if self.config['dataset']["img_norm"]:
test_features = normalize(test_features, dim=0, p=2)
# 对 question 转化为 index
question_length = len(test_question.split())
q_indices = self.que_vocabulary.to_indices(test_question.split())
test_question = self.pad_sequences(q_indices)
item = {}
item['id'] = test_id # scalar
item['question'] = test_question # (max_len,)
item['question_length'] = question_length # scalar
item['features'] = test_features # (36,2048)
item['img_relations'] = img_relations # (36,36,7)
item['facts_num_nodes'] = len(test_top_facts['nodes']) # scalar
item['facts_features'] = test_top_facts['features'] # (num,2048)
item['facts_e1ids'] = test_top_facts['e1ids'] # (num_edges,)
item['facts_e2ids'] = test_top_facts['e2ids'] # (num_edges,)
item['facts_answer'] = test_top_facts['answer']
item['facts_answer_id'] = test_top_facts['answer_id'] # scalar
item['semantic_num_nodes'] = len(semantic_graph['nodes'])
item['semantic_n_features'] = semantic_graph['n_features']
item['semantic_e1ids'] = semantic_graph['e1ids']
item['semantic_e2ids'] = semantic_graph['e2ids']
item['semantic_e_features'] = semantic_graph['e_features']
return item
def __len__(self):
return len(self.test_qids)
def pad_sequences(self, sequence):
sequence = sequence[:self.config['dataset']['max_sequence_lengtn']]
padding = np.zeros(self.config['dataset']['max_sequence_lengtn'])
padding[:len(sequence)] = np.array(sequence)
return torch.tensor(padding)
class FvqaTrainDataset(Dataset):
def __init__(self, config, overfit=False):
super().__init__()
self.que_vocabulary = Vocabulary(config["dataset"]["word2id_path"])
self.image_features = []
self.image_boxes = []
self.image_labels = []
self.image_captions = []
self.questions = []
self.questions_ids = []
self.questions_vecs = []
self.answers = []
self.config = config
# 问题的 id
with open(config['dataset']['train']['train_qids'], 'r') as f:
self.train_qids = json.load(f)
# 问题
with open(config['dataset']['train']['train_questions'], 'r') as f:
self.train_questions = json.load(f)
# 答案
with open(config['dataset']['train']['train_answers'], 'r') as f:
self.train_answers = json.load(f)
# gt fact
with open(config['dataset']['train']['train_gt_facts'], 'r') as f:
self.train_gt_facts = json.load(f)
# caption
with open(config['dataset']['train']['train_captions'], 'r') as f:
self.train_captions = json.load(f)
# semantic graph
with open(config['dataset']['train']['train_semantic_graph'], 'rb') as f:
self.semantic_graphs = pickle.load(f, encoding='iso-8859-1')
# 图像bbox 对应的 label
with open(config['dataset']['train']['train_labels'], 'r') as f:
self.train_labels = json.load(f)
# 图像的长宽
with open(config['dataset']['train']['train_whs'], 'r') as f:
self.train_whs = json.load(f)
# 抽取到的 facts
# with open(config['dataset']['train']['train_facts_graph'], 'rb') as f:
# self.train_facts = pickle.load(f, encoding='iso-8859-1')
with open(config['dataset']['train']['train100_facts_graph'],
'rb') as f:
self.train_top_facts = pickle.load(f, encoding='iso-8859-1')
# 图像 bbox 的特征
self.train_features = np.load(
config['dataset']['train']['train_features'])
# 图像的 bbox 几何信息
self.train_bboxes = np.load(config['dataset']['train']['train_bboxes'])
if overfit:
self.train_qids = self.train_qids[:100]
self.train_questions = self.train_questions[:100]
self.train_answers = self.train_answers[:100]
self.train_gt_facts = self.train_gt_facts[:100]
# self.train_facts = self.train_facts[:100]
self.train_top_facts = self.train_top_facts[:100]
self.train_captions = self.train_captions[:100]
self.train_bboxes = self.train_bboxes[:100]
self.train_features = self.train_features[:100]
self.train_whs = self.train_whs[:100]
self.semantic_graphs=self.semantic_graphs[:100]
def __getitem__(self, index):
train_id = self.train_qids[index]
train_question = self.train_questions[index]
train_answer = self.train_answers[index]
train_gt_fact = self.train_gt_facts[index]
# train_facts = self.train_facts[index]
train_top_facts = self.train_top_facts[index]
train_captions = self.train_captions[index]
train_bboxes = self.train_bboxes[index] # (36,4)
train_features = torch.tensor(self.train_features[index])
semantic_graph=self.semantic_graphs[index]
# cal relation info （36,36,7）
w = self.train_whs[index][0]
h = self.train_whs[index][1]
img_relations = torch.zeros(36, 36, 7)
for i in range(36):
for j in range(36):
xi = train_bboxes[i][0]
yi = train_bboxes[i][1]
wi = train_bboxes[i][2]
hi = train_bboxes[i][3]
xj = train_bboxes[j][0]
yj = train_bboxes[j][1]
wj = train_bboxes[j][2]
hj = train_bboxes[j][3]
r1 = (xj - xi) / (wi * hi) ** 0.5
r2 = (yj - yi) / (wi * hi) ** 0.5
r3 = wj / wi
r4 = hj / hi
r5 = (wj * hj) / wi * hi
r6 = sqrt((xj - xi) ** 2 + (yj - yi) ** 2) / sqrt(w ** 2 + h ** 2)
r7 = atan2(yj - yi, xj - xi)
rel = torch.tensor([r1, r2, r3, r4, r5, r6, r7])
img_relations[i][j] = rel
# 归一化
if self.config['dataset']["img_norm"]:
train_features = normalize(train_features, dim=0, p=2)
# 对 question 转化为 index
question_length = len(train_question.split())
# train_question=self.que_vocabulary.to_indices(train_question.split())
q_indices = self.que_vocabulary.to_indices(train_question.split())
train_question = self.pad_sequences(q_indices)
item = {}
# question
item['id'] = train_id # scalar
item['question'] = train_question # (max_len,)
# item['gt_fact'] = train_gt_fact # [e1,e2,r]
# item['answer'] = train_answer # string
item['question_length'] = question_length # scalar
# image
item['features'] = train_features # (36,2048)
# item['bboxes'] = train_bboxes # (36,4)
# item['captions'] = train_captions # ()
item['img_relations'] = img_relations # (36,36,7)
# fact graph
item['facts_num_nodes'] = len(train_top_facts['nodes']) # scalar
# item['facts_nodes'] = train_top_facts['nodes'] # (num_nodes,)
item['facts_features'] = train_top_facts['features'] # (num,2048)
item['facts_e1ids'] = train_top_facts['e1ids'] # (num_edges,)
item['facts_e2ids'] = train_top_facts['e2ids'] # (num_edges,)
# (num_nodes,) one-hot
item['facts_answer'] = train_top_facts['answer']
item['facts_answer_id'] = train_top_facts['answer_id'] # scalar
item['semantic_num_nodes']=len(semantic_graph['nodes'])
item['semantic_n_features']=semantic_graph['n_features']
item['semantic_e1ids']=semantic_graph['e1ids']
item['semantic_e2ids']=semantic_graph['e2ids']
item['semantic_e_features']=semantic_graph['e_features']
return item
def __len__(self):
return len(self.train_qids)
def pad_sequences(self, sequence):
# 超出的裁剪
sequence = sequence[:self.config['dataset']['max_sequence_lengtn']]
# 没超出的padding
padding = np.zeros(self.config['dataset']['max_sequence_lengtn'])
padding[:len(sequence)] = np.array(sequence)
return torch.tensor(padding)
