vqa-lxmert.py

# coding=utf-8
# Copyright 2020 The HuggingFace Datasets Authors and the current dataset script contributor.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""The Visual Question Answering (VQA) dataset preprocessed for LXMERT."""

import base64
import csv
import json
import os
import sys

import datasets
import numpy as np


csv.field_size_limit(sys.maxsize)


_CITATION = """\
@inproceedings{antol2015vqa,
  title={Vqa: Visual question answering},
  author={Antol, Stanislaw and Agrawal, Aishwarya and Lu, Jiasen and Mitchell, Margaret and Batra, Dhruv and Zitnick, C Lawrence and Parikh, Devi},
  booktitle={Proceedings of the IEEE international conference on computer vision},
  pages={2425--2433},
  year={2015}
}
"""

_DESCRIPTION = """\
VQA is a new dataset containing open-ended questions about images. 
These questions require an understanding of vision, language and commonsense knowledge to answer.
"""

_URLS = {
    "train": "https://nlp.cs.unc.edu/data/lxmert_data/vqa/train.json",
    "train_feat": "https://nlp.cs.unc.edu/data/lxmert_data/mscoco_imgfeat/train2014_obj36.zip",
    "dev": "https://nlp.cs.unc.edu/data/lxmert_data/vqa/valid.json",
    "dev_feat": "https://nlp.cs.unc.edu/data/lxmert_data/mscoco_imgfeat/val2014_obj36.zip",
    "ans2label": "https://raw.githubusercontent.com/airsplay/lxmert/master/data/vqa/trainval_ans2label.json",
}

_TRAIN_IMG_PATH = "train2014_obj36.tsv"
_DEV_IMG_PATH = "mscoco_imgfeat/val2014_obj36.tsv"

FIELDNAMES = [
    "img_id", "img_h", "img_w", "objects_id", "objects_conf", "attrs_id", "attrs_conf", "num_boxes", "boxes", "features"
]

_SHAPE_FEATURES = (36, 2048)
_SHAPE_BOXES = (36, 4)


class VqaV2Lxmert(datasets.GeneratorBasedBuilder):
    """The VQAv2.0 dataset preprocessed for LXMERT, with the objects features detected by a Faster RCNN replacing the
    raw images."""

    BUILDER_CONFIGS = [
        datasets.BuilderConfig(name="vqa", version=datasets.Version("2.0.0"), description="VQA version 2 dataset."),
    ]

    def _info(self):
        features = datasets.Features(
            {
                "question": datasets.Value("string"),
                "question_type": datasets.Value("string"),
                "question_id": datasets.Value("int32"),
                "image_id": datasets.Value("string"),
                "features": datasets.Array2D(_SHAPE_FEATURES, dtype="float32"),
                "boxes": datasets.Array2D(_SHAPE_BOXES, dtype="float32"),
                "answer_type": datasets.Value("string"),
                "label": datasets.Sequence(
                    {
                        "ids": datasets.Value("int32"),
                        "weights": datasets.Value("float32"),
                    }
                ),
            }
        )
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=features,
            supervised_keys=None,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        """Returns SplitGenerators."""
        dl_dir = dl_manager.download_and_extract(_URLS)
        self.ans2label = json.load(open(dl_dir["ans2label"]))

        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={"filepath": dl_dir["train"], "imgfeat": os.path.join(dl_dir["train_feat"], _TRAIN_IMG_PATH)},
            ),
            datasets.SplitGenerator(
                name=datasets.Split.VALIDATION,
                gen_kwargs={"filepath": dl_dir["dev"], "imgfeat": os.path.join(dl_dir["dev_feat"], _DEV_IMG_PATH)},
            ),
        ]

    def _load_features(self, filepath):
        """Returns a dictionary mapping an image id to the corresponding image's objects features."""
        id2features = {}
        with open(filepath) as f:
            reader = csv.DictReader(f, FIELDNAMES, delimiter="\t")
            for i, item in enumerate(reader):
                features = {}
                for key in ["img_h", "img_w", "num_boxes"]:
                    features[key] = int(item[key])
                num_boxes = features["num_boxes"]
                decode_config = [
                    ("objects_id", (num_boxes,), np.int64),
                    ("objects_conf", (num_boxes,), np.float32),
                    ("attrs_id", (num_boxes,), np.int64),
                    ("attrs_conf", (num_boxes,), np.float32),
                    ("boxes", (num_boxes, 4), np.float32),
                    ("features", (num_boxes, -1), np.float32),
                ]
                for key, shape, dtype in decode_config:
                    features[key] = np.frombuffer(base64.b64decode(item[key]), dtype=dtype).reshape(shape)
                id2features[item["img_id"]] = features
        return id2features

    def _generate_examples(self, filepath, imgfeat):
        """ Yields examples as (key, example) tuples."""
        id2features = self._load_features(imgfeat)
        with open(filepath, encoding="utf-8") as f:
            vqa = json.load(f)
            for id_, d in enumerate(vqa):
                img_features = id2features[d["img_id"]]
                ids = [self.ans2label[x] for x in d["label"].keys()]
                weights = list(d["label"].values())
                yield id_, {
                    "question": d["sent"],
                    "question_type": d["question_type"],
                    "question_id": d["question_id"],
                    "image_id": d["img_id"],
                    "features": img_features["features"],
                    "boxes": img_features["boxes"],
                    "answer_type": d["answer_type"],
                    "label": {
                        "ids": ids,
                        "weights": weights,
                    },
                }