wordpress title description,惠州seo排名公司,阿里巴巴运营,洮南网站12种猫分类比赛传送门
要求很简单#xff0c;给train和test集#xff0c;训练模型实现图像分类。
这里使用的是残差连接模型#xff0c;这个平台有预训练好的模型#xff0c;可以直接拿来主义。
训练十几个迭代#xff0c;每个批次60左右#xff0c;准确率达到90%以上…12种猫分类比赛传送门
要求很简单给train和test集训练模型实现图像分类。
这里使用的是残差连接模型这个平台有预训练好的模型可以直接拿来主义。
训练十几个迭代每个批次60左右准确率达到90%以上
一、导入库解压文件 import os
import zipfile
import random
import json
import cv2
import numpy as np
from PIL import Imageimport matplotlib.pyplot as plt
from sklearn.model_selection import train_test_split
import paddle
import paddle.nn as nn
from paddle.io import Dataset,DataLoader
from paddle.nn import \Layer, \Conv2D, Linear, \Embedding, MaxPool2D, \BatchNorm2D, ReLUimport paddle.vision.transforms as transforms
from paddle.vision.models import resnet50
from paddle.metric import Accuracytrain_parameters {input_size: [3, 224, 224], # 输入图片的shapeclass_dim: 12, # 分类数src_path:data/data10954/cat_12_train.zip, # 原始数据集路径src_test_path:data/data10954/cat_12_test.zip, # 原始数据集路径target_path:/home/aistudio/data/dataset, # 要解压的路径 train_list_path: ./train.txt, # train_data.txt路径eval_list_path: ./eval.txt, # eval_data.txt路径label_dict:{}, # 标签字典readme_path: /home/aistudio/data/readme.json,# readme.json路径num_epochs:6, # 训练轮数train_batch_size: 16, # 批次的大小learning_strategy: { # 优化函数相关的配置lr: 0.0005 # 超参数学习率}
}scr_pathtrain_parameters[src_path]
target_pathtrain_parameters[target_path]
src_test_pathtrain_parameters[src_test_path]
z zipfile.ZipFile(scr_path, r)
z.extractall(pathtarget_path)
z zipfile.ZipFile(src_test_path, r)
z.extractall(pathtarget_path)
z.close()
for imgpath in os.listdir(target_path /cat_12_train):src os.path.join(target_path /cat_12_train/, imgpath)img Image.open(src)if img.mode ! RGB:img img.convert(RGB)img.save(src)for imgpath in os.listdir(target_path /cat_12_test):src os.path.join(target_path /cat_12_test/, imgpath)img Image.open(src)if img.mode ! RGB:img img.convert(RGB)img.save(src)解压后将所有图像变为RGB图像
二、加载训练集进行预处理、数据增强、格式变换 transform transforms.Compose([transforms.Resize(size224),transforms.ColorJitter(0.2, 0.2, 0.2, 0.2),transforms.RandomHorizontalFlip(),transforms.RandomRotation(15),transforms.RandomResizedCrop(size224, scale(0.8, 1.0)),transforms.ToTensor(),transforms.Normalize(mean[0.485, 0.456, 0.406], std[0.229, 0.224, 0.225])
])x_train,x_eval,y_train[],[],[]#获取训练图像和标签、测试图像和标签
contents[]
with open(data/data10954/train_list.txt)as f:contentsf.read().split(\n)for item in contents:if item:continuepathdata/dataset/item.split(\t)[0]datanp.array(Image.open(path).convert(RGB))datanp.array(transform(data))x_train.append(data)y_train.append(int(item.split(\t)[-1]))contetnsos.listdir(data/dataset/cat_12_test)
for item in contetns:pathdata/dataset/cat_12_test/itemdatanp.array(Image.open(path).convert(RGB))datanp.array(transform(data))x_eval.append(data)重点是transforms变换的预处理
三、划分训练集和测试集 x_trainnp.array(x_train)y_trainnp.array(y_train)x_evalnp.array(x_eval)x_train,x_test,y_train,y_testtrain_test_split(x_train,y_train,test_size0.2,random_state42,stratifyy_train)x_trainpaddle.to_tensor(x_train,dtypefloat32)
y_trainpaddle.to_tensor(y_train,dtypeint64)
x_testpaddle.to_tensor(x_test,dtypefloat32)
y_testpaddle.to_tensor(y_test,dtypeint64)
x_evalpaddle.to_tensor(x_eval,dtypefloat32) 这是必要的可以随时利用测试集查看准确率
四、加载预训练模型选择损失函数和优化器 learning_rate0.001
epochs 5 # 迭代轮数
batch_size 50 # 批次大小
weight_decay1e-5
num_class12cnnresnet50(pretrainedTrue)
checkpointpaddle.load(checkpoint.pdparams)for param in cnn.parameters():param.requires_gradFalse
cnn.fc nn.Linear(2048, num_class)
cnn.set_dict(checkpoint[cnn_state_dict])
criterionnn.CrossEntropyLoss()
optimizer paddle.optimizer.Adam(learning_ratelearning_rate, parameterscnn.fc.parameters(),weight_decayweight_decay) 第一次训练把加载模型注释掉即可优化器包含最后一层全连接的参数
五、模型训练 if x_train.shape[3]3:x_trainpaddle.transpose(x_train,perm(0,3,1,2))dataset paddle.io.TensorDataset([x_train, y_train])
data_loader DataLoader(dataset, batch_sizebatch_size, shuffleTrue)
for epoch in range(epochs):for batch_data, batch_labels in data_loader:outputs cnn(batch_data)loss criterion(outputs, batch_labels)print(epoch)loss.backward()optimizer.step()optimizer.clear_grad()print(fEpoch [{epoch1}/{epochs}], Loss: {loss.numpy()[0]})#保存参数
paddle.save({cnn_state_dict: cnn.state_dict(),}, checkpoint.pdparams)使用批处理这个很重要不然平台分分钟炸了
六、测试集准确率 num_class12
batch_size64
cnnresnet50(pretrainedTrue)
checkpointpaddle.load(checkpoint.pdparams)for param in cnn.parameters():param.requires_gradFalse
cnn.fc nn.Linear(2048, num_class)
cnn.set_dict(checkpoint[cnn_state_dict])cnn.eval()if x_test.shape[3]3:x_testpaddle.transpose(x_test,perm(0,3,1,2))
dataset paddle.io.TensorDataset([x_test, y_test])
data_loader DataLoader(dataset, batch_sizebatch_size, shuffleTrue)with paddle.no_grad():score0for batch_data, batch_labels in data_loader:predictions cnn(batch_data)predicted_probabilities paddle.nn.functional.softmax(predictions, axis1)predicted_labels paddle.argmax(predicted_probabilities, axis1) print(predicted_labels)for i in range(len(predicted_labels)):if predicted_labels[i].numpy()batch_labels[i]:score1print(score/len(y_test)) 设置eval模式使用批处理测试准确率
