def load_item(self, idx):
sample_info = self.annotation_db[idx]
sample_info = self.preprocess_sample_info(sample_info)
current_sample = Sample()
if self._dataset_type != "test":
text_processor_argument = {"tokens": sample_info["caption_tokens"]}
processed_caption = self.text_processor(text_processor_argument)
current_sample.text = processed_caption["text"]
current_sample.caption_id = torch.tensor(sample_info["caption_id"],
dtype=torch.int)
current_sample.caption_len = torch.tensor(len(
sample_info["caption_tokens"]),
dtype=torch.int)
current_sample.image_id = object_to_byte_tensor(
sample_info["image_id"])
if self._use_features:
features = self.features_db[idx]
current_sample.update(features)
else:
image_path = str(sample_info["image_name"]) + ".jpg"
current_sample.image = self.image_db.from_path(
image_path)["images"][0]
# Add reference captions to sample
current_sample = self.add_reference_caption(sample_info,
current_sample)
return current_sample
def load_item(self, idx):
sample_info = self.annotation_db[idx]
current_sample = Sample()
processed_caption = self.text_processor(
{"text": sample_info["captions"][0]})
current_sample.text = processed_caption["text"]
current_sample.caption_len = torch.tensor(len(
processed_caption["text"]),
dtype=torch.int)
if isinstance(sample_info["image_id"], int):
current_sample.image_id = torch.tensor(sample_info["image_id"],
dtype=torch.int)
else:
current_sample.image_id = sample_info["image_id"]
if self._use_features is True:
features = self.features_db[idx]
current_sample.update(features)
current_sample.answers = torch.stack([processed_caption["text"]])
return current_sample