svhn

הפניות:

• קוד
• מחבקת פנים

מלא_מספרים

השתמש בפקודה הבאה כדי לטעון מערך נתונים זה ב-TFDS:

ds = tfds.load('huggingface:svhn/full_numbers')

• תיאור :

SVHN is a real-world image dataset for developing machine learning and object recognition algorithms with minimal requirement on data preprocessing and formatting.
It can be seen as similar in flavor to MNIST (e.g., the images are of small cropped digits), but incorporates an order of magnitude more labeled data (over 600,000 digit images)
and comes from a significantly harder, unsolved, real world problem (recognizing digits and numbers in natural scene images). SVHN is obtained from house numbers in Google Street View images.

• רישיון : מותאם אישית (לא מסחרי)
• גרסה : 1.0.0
• פיצולים :

• תכונות :

{
    "image": {
        "id": null,
        "_type": "Image"
    },
    "digits": {
        "feature": {
            "bbox": {
                "feature": {
                    "dtype": "int32",
                    "id": null,
                    "_type": "Value"
                },
                "length": 4,
                "id": null,
                "_type": "Sequence"
            },
            "label": {
                "num_classes": 10,
                "names": [
                    "0",
                    "1",
                    "2",
                    "3",
                    "4",
                    "5",
                    "6",
                    "7",
                    "8",
                    "9"
                ],
                "names_file": null,
                "id": null,
                "_type": "ClassLabel"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

ספרות_חתוכות

השתמש בפקודה הבאה כדי לטעון מערך נתונים זה ב-TFDS:

ds = tfds.load('huggingface:svhn/cropped_digits')

• תיאור :

SVHN is a real-world image dataset for developing machine learning and object recognition algorithms with minimal requirement on data preprocessing and formatting.
It can be seen as similar in flavor to MNIST (e.g., the images are of small cropped digits), but incorporates an order of magnitude more labeled data (over 600,000 digit images)
and comes from a significantly harder, unsolved, real world problem (recognizing digits and numbers in natural scene images). SVHN is obtained from house numbers in Google Street View images.

• רישיון : מותאם אישית (לא מסחרי)
• גרסה : 1.0.0
• פיצולים :

• תכונות :

{
    "image": {
        "id": null,
        "_type": "Image"
    },
    "label": {
        "num_classes": 10,
        "names": [
            "0",
            "1",
            "2",
            "3",
            "4",
            "5",
            "6",
            "7",
            "8",
            "9"
        ],
        "names_file": null,
        "id": null,
        "_type": "ClassLabel"
    }
}