[
    [
        {
            "type": "image",
            "content": {
                "image_source": {
                    "path": "images/a20c38a3eb7c81159ade9746e04e47b843e6e947dfc07103cf0406c615dd044e.jpg"
                },
                "content": "",
                "image_caption": [
                    {
                        "type": "text",
                        "content": "Fig. 2 The ILSVRC dataset contains many more fine-grained classes compared to the standard PASCAL VOC benchmark; for example, instead of the PASCAL “dog” category there are 120 diferent breeds of dogs in ILSVRC2012-2014 classification and single-object localization tasks."
                    }
                ],
                "image_footnote": []
            },
            "bbox": [
                117,
                102,
                823,
                381
            ]
        },
        {
            "type": "paragraph",
            "content": {
                "paragraph_content": [
                    {
                        "type": "text",
                        "content": "are 1000 object classes and approximately 1.2 million training images, 50 thousand validation images and 100 thousand test images. Table 2 (top) documents the size of the dataset over the years of the challenge."
                    }
                ]
            },
            "bbox": [
                63,
                460,
                473,
                521
            ]
        },
        {
            "type": "title",
            "content": {
                "title_content": [
                    {
                        "type": "text",
                        "content": "3.2 Single-object localization dataset construction"
                    }
                ],
                "level": 2
            },
            "bbox": [
                65,
                545,
                438,
                560
            ]
        },
        {
            "type": "paragraph",
            "content": {
                "paragraph_content": [
                    {
                        "type": "text",
                        "content": "The single-object localization task evaluates the ability of an algorithm to localize one instance of an object category. It was introduced as a taster task in ILSVRC 2011, and became an oficial part of ILSVRC in 2012."
                    }
                ]
            },
            "bbox": [
                63,
                571,
                472,
                632
            ]
        },
        {
            "type": "paragraph",
            "content": {
                "paragraph_content": [
                    {
                        "type": "text",
                        "content": "The key challenge was developing a scalable crowdsourcing method for object bounding box annotation. Our three-step self-verifying pipeline is described in Section 3.2.1. Having the dataset collected, we perform detailed analysis in Section 3.2.2 to ensure that the dataset is suficiently varied to be suitable for evaluation of object localization algorithms."
                    }
                ]
            },
            "bbox": [
                63,
                633,
                473,
                737
            ]
        },
        {
            "type": "paragraph",
            "content": {
                "paragraph_content": [
                    {
                        "type": "text",
                        "content": "Object classes and candidate images. The object classes for single-object localization task are the same as the object classes for image classification task described above in Section 3.1. The training images for localization task are a subset of the training images used for image classification task, and the validation and test images are the same between both tasks."
                    }
                ]
            },
            "bbox": [
                63,
                750,
                473,
                854
            ]
        },
        {
            "type": "paragraph",
            "content": {
                "paragraph_content": [
                    {
                        "type": "text",
                        "content": "Bounding box annotation. Recall that for the image classification task every image was annotated with one object class label, corresponding to one object that is present in an image. For the single-object localization task, every validation and test image and a subset of the training images are annotated with axis-aligned bounding boxes around every instance of this object."
                    }
                ]
            },
            "bbox": [
                63,
                866,
                473,
                897
            ]
        },
        {
            "type": "paragraph",
            "content": {
                "paragraph_content": []
            },
            "bbox": [
                492,
                460,
                902,
                533
            ]
        },
        {
            "type": "paragraph",
            "content": {
                "paragraph_content": [
                    {
                        "type": "text",
                        "content": "Every bounding box is required to be as small as possible while including all visible parts of the object instance. An alternate annotation procedure could be to annotate the full (estimated) extent of the object: e.g., if a person’s legs are occluded and only the torso is visible, the bounding box could be drawn to include the likely location of the legs. However, this alternative procedure is inherently ambiguous and ill-defined, leading to disagreement among annotators and among researchers (what is the true “most likely” extent of this object?). We follow the standard protocol of only annotating visible object parts (Russell et al., 2007; Everingham et al., 2010).<sup>5</sup>"
                    }
                ]
            },
            "bbox": [
                492,
                535,
                902,
                728
            ]
        },
        {
            "type": "title",
            "content": {
                "title_content": [
                    {
                        "type": "text",
                        "content": "3.2.1 Bounding box object annotation system"
                    }
                ],
                "level": 2
            },
            "bbox": [
                494,
                743,
                828,
                758
            ]
        },
        {
            "type": "paragraph",
            "content": {
                "paragraph_content": [
                    {
                        "type": "text",
                        "content": "We summarize the crowdsourced bounding box annotation system described in detail in (Su et al., 2012). The goal is to build a system that is fully automated,"
                    }
                ]
            },
            "bbox": [
                492,
                771,
                902,
                816
            ]
        },
        {
            "type": "page_number",
            "content": {
                "page_number_content": [
                    {
                        "type": "text",
                        "content": "8"
                    }
                ]
            },
            "bbox": [
                67,
                78,
                80,
                89
            ]
        },
        {
            "type": "page_header",
            "content": {
                "page_header_content": [
                    {
                        "type": "text",
                        "content": "Olga Russakovsky* et al."
                    }
                ]
            },
            "bbox": [
                729,
                78,
                899,
                90
            ]
        },
        {
            "type": "page_footnote",
            "content": {
                "page_footnote_content": [
                    {
                        "type": "text",
                        "content": "<sup>5</sup> Some datasets such as PASCAL VOC (Everingham et al., 2010) and LabelMe (Russell et al., 2007) are able to provide more detailed annotations: for example, marking individual object instances as being truncated. We chose not to provide this level of detail in favor of annotating more images and more object instances."
                    }
                ]
            },
            "bbox": [
                492,
                822,
                902,
                897
            ]
        }
    ]
]