{ "id": 244045, "title": "请问怎么定义target_transform将标签改成顺序型数据？", "url": "/competitions/classify-leaves/discussion/244045", "author_name": "", "post_date": "2021-06-05T02:58:44.767902600Z", "votes": 1, "comment_count": 4, "views": 0, "content": "

我定义了一个Dataset类来生成DataLoader，请问怎么用target_transform将标签（字符串）改成顺序型数据。下面是改成one hot的示例。

\n

from torchvision import datasets
\nfrom torchvision.transforms import ToTensor, Lambda

\n

ds = datasets.FashionMNIST(
\n root=\"data\",
\n train=True,
\n download=True,
\n transform=ToTensor(),
\n target_transform=Lambda(lambda y: torch.zeros(10, dtype=torch.float).scatter_(0, torch.tensor(y), value=1))
\n)

", "messages": [ { "id": "1336541", "postDate": "06/05/2021 02:58:44", "content": "

我定义了一个Dataset类来生成DataLoader，请问怎么用target_transform将标签（字符串）改成顺序型数据。下面是改成one hot的示例。

\n

from torchvision import datasets
\nfrom torchvision.transforms import ToTensor, Lambda

\n

ds = datasets.FashionMNIST(
\n root=\"data\",
\n train=True,
\n download=True,
\n transform=ToTensor(),
\n target_transform=Lambda(lambda y: torch.zeros(10, dtype=torch.float).scatter_(0, torch.tensor(y), value=1))
\n)

", "rawMarkdown": "我定义了一个Dataset类来生成DataLoader，请问怎么用target_transform将标签（字符串）改成顺序型数据。下面是改成one hot的示例。\n\nfrom torchvision import datasets\nfrom torchvision.transforms import ToTensor, Lambda\n\nds = datasets.FashionMNIST(\n root=\"data\",\n train=True,\n download=True,\n transform=ToTensor(),\n target_transform=Lambda(lambda y: torch.zeros(10, dtype=torch.float).scatter_(0, torch.tensor(y), value=1))\n)", "votes": null }, { "id": "1341795", "postDate": "06/09/2021 02:35:31", "content": "

可以参考以下方法对string类型标签进行编码为Int型类别，当然也可以继续转换成ont hot类型

\n
train_df = pd.read_csv('path/to/train.csv')\nlabels = train_df.label.unique()\nlabel_encoder = {}\nfor idx, name in enumerate(labels):\n    label_encoder.update({name:idx})\n\ntrain_df.label = train_df.label.apply(lambda x: label_encoder[x])\n
", "rawMarkdown": "可以参考以下方法对string类型标签进行编码为Int型类别，当然也可以继续转换成ont hot类型\n```python\ntrain_df = pd.read_csv('path/to/train.csv')\nlabels = train_df.label.unique()\nlabel_encoder = {}\nfor idx, name in enumerate(labels):\n label_encoder.update({name:idx})\n\ntrain_df.label = train_df.label.apply(lambda x: label_encoder[x])\n```", "votes": null }, { "id": "1345947", "postDate": "06/12/2021 01:44:57", "content": "

我的方法是对所有标签去重后排序，建立一个映射之后getitem中将每个item的label对应上映射中的index

", "rawMarkdown": "我的方法是对所有标签去重后排序，建立一个映射之后__getitem__中将每个item的label对应上映射中的index", "votes": null }, { "id": "1354190", "postDate": "06/17/2021 12:45:58", "content": "

https://www.kaggle.com/nekokiku/simple-resnet-baseline
\n参考这位同学提供的baseline，可以说是很详细了。👍

", "rawMarkdown": "https://www.kaggle.com/nekokiku/simple-resnet-baseline\n参考这位同学提供的baseline，可以说是很详细了。👍", "votes": null }, { "id": "1367713", "postDate": "06/28/2021 03:46:22", "content": "

df = pd.read_csv('mount/My Drive/Colab Notebooks/d2l/data/classify-leaves/train.csv')
\ncategory = df[\"label\"].unique()
\ndict = {}
\nfor i,c in enumerate(category):
\n dict[c] = i

\n

target_transform = transforms.Lambda(lambda y:dict[y])

", "rawMarkdown": "df = pd.read_csv('mount/My Drive/Colab Notebooks/d2l/data/classify-leaves/train.csv')\ncategory = df[\"label\"].unique()\ndict = {}\nfor i,c in enumerate(category):\n dict[c] = i\n\ntarget_transform = transforms.Lambda(lambda y:dict[y])", "votes": null } ], "comments": [ { "id": 1341795, "author_name": "seefun", "author_url": "", "post_date": "06/09/2021 02:35:31", "content": "

可以参考以下方法对string类型标签进行编码为Int型类别，当然也可以继续转换成ont hot类型

\n
train_df = pd.read_csv('path/to/train.csv')\nlabels = train_df.label.unique()\nlabel_encoder = {}\nfor idx, name in enumerate(labels):\n    label_encoder.update({name:idx})\n\ntrain_df.label = train_df.label.apply(lambda x: label_encoder[x])\n
", "votes": null, "replies": [] }, { "id": 1345947, "author_name": "weishuozhao", "author_url": "", "post_date": "06/12/2021 01:44:57", "content": "

我的方法是对所有标签去重后排序，建立一个映射之后getitem中将每个item的label对应上映射中的index

", "votes": null, "replies": [] }, { "id": 1354190, "author_name": "charlesyyun", "author_url": "", "post_date": "06/17/2021 12:45:58", "content": "

https://www.kaggle.com/nekokiku/simple-resnet-baseline
\n参考这位同学提供的baseline，可以说是很详细了。👍

", "votes": null, "replies": [] }, { "id": 1367713, "author_name": "liujiedeeplearning", "author_url": "", "post_date": "06/28/2021 03:46:22", "content": "

df = pd.read_csv('mount/My Drive/Colab Notebooks/d2l/data/classify-leaves/train.csv')
\ncategory = df[\"label\"].unique()
\ndict = {}
\nfor i,c in enumerate(category):
\n dict[c] = i

\n

target_transform = transforms.Lambda(lambda y:dict[y])

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1336541": "我定义了一个Dataset类来生成DataLoader，请问怎么用target_transform将标签（字符串）改成顺序型数据。下面是改成one hot的示例。\n\nfrom torchvision import datasets\nfrom torchvision.transforms import ToTensor, Lambda\n\nds = datasets.FashionMNIST(\n root=\"data\",\n train=True,\n download=True,\n transform=ToTensor(),\n target_transform=Lambda(lambda y: torch.zeros(10, dtype=torch.float).scatter_(0, torch.tensor(y), value=1))\n)", "1341795": "可以参考以下方法对string类型标签进行编码为Int型类别，当然也可以继续转换成ont hot类型\n```python\ntrain_df = pd.read_csv('path/to/train.csv')\nlabels = train_df.label.unique()\nlabel_encoder = {}\nfor idx, name in enumerate(labels):\n label_encoder.update({name:idx})\n\ntrain_df.label = train_df.label.apply(lambda x: label_encoder[x])\n```", "1345947": "我的方法是对所有标签去重后排序，建立一个映射之后__getitem__中将每个item的label对应上映射中的index", "1354190": "https://www.kaggle.com/nekokiku/simple-resnet-baseline\n参考这位同学提供的baseline，可以说是很详细了。👍", "1367713": "df = pd.read_csv('mount/My Drive/Colab Notebooks/d2l/data/classify-leaves/train.csv')\ncategory = df[\"label\"].unique()\ndict = {}\nfor i,c in enumerate(category):\n dict[c] = i\n\ntarget_transform = transforms.Lambda(lambda y:dict[y])" }, "source": "meta" }