{
    "info": {
        "author": "Maxim Bonnaerens",
        "author_email": "maxim@bonnaerens.be",
        "bugtrack_url": null,
        "classifiers": [],
        "description": "# Visual Wake Words Dataset\nPython library to work with the [Visual Wake Words Dataset](https://arxiv.org/abs/1906.05721), \ncomparable to [pycococools](https://github.com/cocodataset/cocoapi) for the COCO dataset.\n\n`pyvww.utils.VisualWakeWords` inherits from `pycocotools.coco.COCO` and can be used in an similar fashion.\n\n`pyvww.pytorch.VisualWakeWordsClassification` is a pytorch `Dataset` which can be used like any \nimage classification dataset.\n\n ---\n ### Installation\n The code is implemented in Python 3.7 and can be installed with pip:\n ```bash\n pip install pyvww\n ```\n\n ### Usage\n The Visual Wake Words Dataset is derived from the publicly available [COCO](cocodataset.org/#/home) dataset.\n To download the COCO dataset use the script `download_coco.sh`\n ```bash\nbash scripts/download_mscoco.sh path-to-COCO-dataset\n```\nThe Visual Wake Words Dataset evaluates the accuracy on the [minival image ids](https://raw.githubusercontent.com/tensorflow/models/master/research/object_detection/data/mscoco_minival_ids.txt),\nand for training uses the remaining 115k images of the COCO training/validation dataset.\n\nTo create COCO annotation files that convert the 83K/41K split to the 115K/8K split use:\n`scripts/create_coco_train_minival_split.py`\n```bash\nTRAIN_ANNOTATIONS_FILE=\"path-to-mscoco-dataset/annotations/instances_train2014.json\"\nVAL_ANNOTATIONS_FILE=\"path-to-mscoco-dataset/annotations/instances_val2014.json\"\nDIR=\"path-to-mscoco-dataset/annotations/\"\npython scripts/create_coco_train_minival_split.py \\\n  --train_annotations_file=\"${TRAIN_ANNOTATIONS_FILE}\" \\\n  --val_annotations_file=\"${VAL_ANNOTATIONS_FILE}\" \\\n--output_dir=\"${DIR}\"\n```\n\nThe process of creating the Visual Wake Words dataset from COCO dataset is as follows.\nEach image is assigned a label 1 or 0. \nThe label 1 is assigned as long as it has at least one bounding box corresponding \nto the object of interest (e.g. person) with the box area greater than a certain threshold \n(e.g. 0.5% of the image area).\n\nTo generate the new annotations, use the script `scripts/create_visualwakewords_annotations.py`.\n```bash\nMAXITRAIN_ANNOTATIONS_FILE=\"path-to-mscoco-dataset/annotations/instances_maxitrain.json\"\nMINIVAL_ANNOTATIONS_FILE=\"path-to-mscoco-dataset/annotations/instances_minival.json\"\nVWW_OUTPUT_DIR=\"new-path-to-visualwakewords-dataset/annotations/\"\npython scripts/create_visualwakewords_annotations.py \\\n  --train_annotations_file=\"${MAXITRAIN_ANNOTATIONS_FILE}\" \\\n  --val_annotations_file=\"${MINIVAL_ANNOTATIONS_FILE}\" \\\n  --output_dir=\"${VWW_OUTPUT_DIR}\" \\\n  --threshold=0.005 \\\n  --foreground_class='person'\n```\n\nThe generated annotations follow the [COCO Data format](http://cocodataset.org/#format-data).\n```\n{\n  \"info\" : info, \n  \"images\" : [image], \n  \"annotations\" : [annotation], \n  \"licenses\" : [license],\n}\n\ninfo{\n  \"year\" : int, \n  \"version\" : str, \n  \"description\" : str, \n  \"url\" : str, \n}\n\nimage{\n  \"id\" : int, \n  \"width\" : int, \n  \"height\" : int, \n  \"file_name\" : str, \n  \"license\" : int, \n  \"flickr_url\" : str, \n  \"coco_url\" : str, \n  \"date_captured\" : datetime,\n}\n\nlicense{\n  \"id\" : int, \n  \"name\" : str, \n  \"url\" : str,\n}\n\nannotation{\n  \"id\" : int, \n  \"image_id\" : int, \n  \"category_id\" : int, \n  \"area\" : float, \n  \"bbox\" : [x,y,width,height], \n  \"iscrowd\" : 0 or 1,\n}\n```\n\n### Pytorch Dataset\n\nThe `pyvww.pytorch.VisualWakeWordsClassification` can be used in pytorch like any other pytorch image classification\ndataset such as MNIST or ImageNet.\n\nNote: If you used the script `create_coco_train_minival_split.py` to create the annotations for the 115k/8k split, \nyou need to move or copy the train2014 and val2014 directories to a shared directory. E.g.:\n```bash\ncd path-to-mscoco-dataset/\nmkdir all\ncp -a train2014/. all/\ncp -a val2014/. all/\n```\n```python\nimport torch\nimport pyvww\n\ntrain_dataset = pyvww.pytorch.VisualWakeWordsClassification(root=\"path-to-mscoco-dataset/all\", \n                    annFile=\".../visualwakewords/annotations/instances_train.json\")\n```\n\n\n\n\n",
        "description_content_type": "text/markdown",
        "docs_url": null,
        "download_url": "",
        "downloads": {
            "last_day": -1,
            "last_month": -1,
            "last_week": -1
        },
        "home_page": "https://github.com/Mxbonn/visualwakewords",
        "keywords": "",
        "license": "Apache 2.0",
        "maintainer": "",
        "maintainer_email": "",
        "name": "pyvww",
        "package_url": "https://pypi.org/project/pyvww/",
        "platform": "",
        "project_url": "https://pypi.org/project/pyvww/",
        "project_urls": {
            "Homepage": "https://github.com/Mxbonn/visualwakewords"
        },
        "release_url": "https://pypi.org/project/pyvww/0.1.0/",
        "requires_dist": [
            "pycocotools"
        ],
        "requires_python": "",
        "summary": "Python API to work with the Visual Wake Words Dataset.",
        "version": "0.1.0"
    },
    "last_serial": 5505307,
    "releases": {
        "0.1.0": [
            {
                "comment_text": "",
                "digests": {
                    "md5": "cf615e78ff992b51dae8f5ba136e6c7e",
                    "sha256": "d1f827901e92b749558f0188b0d02473c822e39584dc1e33ec2a55204815bf41"
                },
                "downloads": -1,
                "filename": "pyvww-0.1.0-py3-none-any.whl",
                "has_sig": false,
                "md5_digest": "cf615e78ff992b51dae8f5ba136e6c7e",
                "packagetype": "bdist_wheel",
                "python_version": "py3",
                "requires_python": null,
                "size": 8964,
                "upload_time": "2019-07-09T07:36:54",
                "url": "https://files.pythonhosted.org/packages/42/ae/ade768da67ab85065a01870a0d09c1615b27e63cd4f1b2135e401e365112/pyvww-0.1.0-py3-none-any.whl"
            },
            {
                "comment_text": "",
                "digests": {
                    "md5": "6933b140ca1114e4863ac58f0a11b87b",
                    "sha256": "2ebed1557f5df29452ec4ff6ca1246060e2f447cc68e50abd81ea1ac0fa054ee"
                },
                "downloads": -1,
                "filename": "pyvww-0.1.0.tar.gz",
                "has_sig": false,
                "md5_digest": "6933b140ca1114e4863ac58f0a11b87b",
                "packagetype": "sdist",
                "python_version": "source",
                "requires_python": null,
                "size": 4162,
                "upload_time": "2019-07-09T07:36:56",
                "url": "https://files.pythonhosted.org/packages/6a/f2/00167f36bee6989d2a321313467d2c2791f3c51938792bed91babf40f8d5/pyvww-0.1.0.tar.gz"
            }
        ]
    },
    "urls": [
        {
            "comment_text": "",
            "digests": {
                "md5": "cf615e78ff992b51dae8f5ba136e6c7e",
                "sha256": "d1f827901e92b749558f0188b0d02473c822e39584dc1e33ec2a55204815bf41"
            },
            "downloads": -1,
            "filename": "pyvww-0.1.0-py3-none-any.whl",
            "has_sig": false,
            "md5_digest": "cf615e78ff992b51dae8f5ba136e6c7e",
            "packagetype": "bdist_wheel",
            "python_version": "py3",
            "requires_python": null,
            "size": 8964,
            "upload_time": "2019-07-09T07:36:54",
            "url": "https://files.pythonhosted.org/packages/42/ae/ade768da67ab85065a01870a0d09c1615b27e63cd4f1b2135e401e365112/pyvww-0.1.0-py3-none-any.whl"
        },
        {
            "comment_text": "",
            "digests": {
                "md5": "6933b140ca1114e4863ac58f0a11b87b",
                "sha256": "2ebed1557f5df29452ec4ff6ca1246060e2f447cc68e50abd81ea1ac0fa054ee"
            },
            "downloads": -1,
            "filename": "pyvww-0.1.0.tar.gz",
            "has_sig": false,
            "md5_digest": "6933b140ca1114e4863ac58f0a11b87b",
            "packagetype": "sdist",
            "python_version": "source",
            "requires_python": null,
            "size": 4162,
            "upload_time": "2019-07-09T07:36:56",
            "url": "https://files.pythonhosted.org/packages/6a/f2/00167f36bee6989d2a321313467d2c2791f3c51938792bed91babf40f8d5/pyvww-0.1.0.tar.gz"
        }
    ]
}