### 实现木薯叶子疾病分类模型 #### 数据准备与预处理 为了构建有效的木薯叶病害分类模型，数据集的质量至关重要。通常情况下，会先下载并解压相应的数据集文件。对于图像数据而言，还需要进行必要的预处理操作，比如调整大小、归一化以及应用各种形式的数据增强技术来扩充训练样本数量。 ```python import os from PIL import Image import numpy as np import torch from torchvision.transforms import Compose, Resize, ToTensor, Normalize from albumentations.pytorch import ToTensorV2 import albumentations as A # 定义图片转换流程 transform = A.Compose([ A.Resize(height=256, width=256), A.Normalize(mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225)), ToTensorV2() ]) def load_image(image_path): img = Image.open(image_path).convert('RGB') transformed_img = transform(image=np.array(img))['image'] return transformed_img.unsqueeze(0) data_dir = 'path_to_dataset' images = [] labels = [] for label_name in sorted(os.listdir(data_dir)): label_dir = os.path.join(data_dir, label_name) if not os.path.isdir(label_dir): continue for image_file in os.listdir(label_dir): images.append(load_image(os.path.join(label_dir, image_file))) labels.append(int(label_name)) ``` #### 构建多模型集成方案 基于之前比赛中的成功经验[^1]，这里采用多个不同架构的神经网络来进行预测，并最终通过对各模型输出概率加权求平均的方式得到综合结果。具体来说，选择了`ResNext50`, `Vision Transformer (ViT)`作为基础组件之一，在后续迭代过程中还可以考虑加入其他类型的卷积神经网络如`EfficientNet-B4`等以进一步提升性能表现。 ```python import timm from transformers import ViTForImageClassification class EnsembleModel(torch.nn.Module): def __init__(self): super().__init__() self.resnext = timm.create_model('resnext50_32x4d', pretrained=True, num_classes=5) self.vit = ViTForImageClassification.from_pretrained('google/vit-base-patch16-224') @staticmethod def forward(model, inputs): outputs = model(**inputs)[0] probs = torch.softmax(outputs, dim=-1) return probs def predict(self, input_tensor): resnext_probs = EnsembleModel.forward(self.resnext, {'pixel_values':input_tensor}) vit_probs = EnsembleModel.forward(self.vit, {'pixel_values':input_tensor}) avg_prob = (resnext_probs + vit_probs)/2. _, predicted_class = torch.max(avg_prob.data, 1) return predicted_class.item(), avg_prob.tolist()[0] ensemble_model = EnsembleModel().eval() with torch.no_grad(): prediction, probabilities = ensemble_model.predict(images[0]) print(f'Predicted class index: {prediction}, Probabilities distribution:{probabilities}') ```