请问关于pytorch版本该如何修改测试自己的数据集？

[chajnoven]

比如自己测试集在“当前目录的test的文件夹”里面

[ypwhs]

class CaptchaDataset(Dataset):
    def __init__(self, characters, length, width, height, input_length, label_length):
        super(CaptchaDataset, self).__init__()
        self.characters = characters
        self.length = length
        self.width = width
        self.height = height
        self.input_length = input_length
        self.label_length = label_length
        self.n_class = len(characters)
        self.generator = ImageCaptcha(width=width, height=height)

    def __len__(self):
        return self.length
    
    def __getitem__(self, index):
        random_str = ''.join([random.choice(self.characters[1:]) for j in range(self.label_length)])
        image = to_tensor(self.generator.generate_image(random_str))
        target = torch.tensor([self.characters.find(x) for x in random_str], dtype=torch.long)
        input_length = torch.full(size=(1, ), fill_value=self.input_length, dtype=torch.long)
        target_length = torch.full(size=(1, ), fill_value=self.label_length, dtype=torch.long)
        return image, target, input_length, target_length

实现一个上面的自定义 Dataset 类， init 改为输入一个文件夹以及必要的信息，然后在 getitem 里读图，只要下面的代码能跑通，你就可以使用此 dataset 去测试。

dataset = CaptchaDataset(characters, 1, width, height, n_input_length, n_len)
image, target, input_length, label_length = dataset[0]
print(''.join([characters[x] for x in target]), input_length, label_length)
to_pil_image(image)

[chajnoven]

非常感谢您的回答，这对于我新手来说是开始，  再次感谢您的帮助 江半德

…

------------------&nbsp;原始邮件&nbsp;------------------ 发件人:&nbsp;"杨培文 (Yang Peiwen)"<[email protected]&gt;; 发送时间:&nbsp;2019年11月1日(星期五) 晚上10:35 收件人:&nbsp;"ypwhs/captcha_break"<[email protected]&gt;; 抄送:&nbsp;"江半德"<[email protected]&gt;;"Author"<[email protected]&gt;; 主题:&nbsp;Re: [ypwhs/captcha_break] 请问关于pytorch版本该如何修改测试自己的数据集？ (#43) class CaptchaDataset(Dataset): def __init__(self, characters, length, width, height, input_length, label_length): super(CaptchaDataset, self).__init__() self.characters = characters self.length = length self.width = width self.height = height self.input_length = input_length self.label_length = label_length self.n_class = len(characters) self.generator = ImageCaptcha(width=width, height=height) def __len__(self): return self.length def __getitem__(self, index): random_str = ''.join([random.choice(self.characters[1:]) for j in range(self.label_length)]) image = to_tensor(self.generator.generate_image(random_str)) target = torch.tensor([self.characters.find(x) for x in random_str], dtype=torch.long) input_length = torch.full(size=(1, ), fill_value=self.input_length, dtype=torch.long) target_length = torch.full(size=(1, ), fill_value=self.label_length, dtype=torch.long) return image, target, input_length, target_length 实现一个上面的自定义 Dataset 类， init 改为输入一个文件夹以及必要的信息，然后在 getitem 里读图，只要下面的代码能跑通，你就可以使用此 dataset 去测试。 dataset = CaptchaDataset(characters, 1, width, height, n_input_length, n_len) image, target, input_length, label_length = dataset[0] print(''.join([characters[x] for x in target]), input_length, label_length) to_pil_image(image) — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.