PaddlePaddle（飞桨）是百度公司开发的深度学习框架，具有易用、高效、灵活和可伸缩等特点。以下是关于PaddlePaddle框架的详细介绍：
- 易用性：PaddlePaddle框架提供了Python作为主要的前端语言，并提供了丰富的编程接口API，方便用户开发和调用模型。同时，PaddlePaddle还提供了大量的预训练模型和工具组件，用户可以直接使用或进行自定义开发。
- 高效性：PaddlePaddle框架采用了高性能的算法和优化技术，支持大规模的模型训练和部署。同时，PaddlePaddle还支持多种硬件加速器，如GPU、FPGA等，进一步提高了计算效率。
- 灵活性：PaddlePaddle框架支持动态图和静态图两种编程模式，用户可以根据需要选择。同时，PaddlePaddle还提供了多种工具和组件，如可视化工具、迁移学习组件等，方便用户进行模型开发和调试。
- 可伸缩性：PaddlePaddle框架支持分布式训练，能够处理大规模的数据集和模型。同时，PaddlePaddle还提供了多种集群训练策略，如数据并行、模型并行等，满足不同场景下的训练需求。
- 生态系统：百度公司为PaddlePaddle框架构建了完整的生态系统，包括开发套件、工具组件、预训练模型库等。同时
基于深度学习的鲜花识别主要涉及以下几个步骤：
- 数据收集与标注：首先，需要收集大量的鲜花图片，并对这些图片进行标注，即确定每张图片上的鲜花种类。这些标注的数据将用于训练深度学习模型。
- 特征提取：使用深度学习技术，特别是卷积神经网络（CNN），对收集的图片进行特征提取。卷积神经网络能够从原始图片中提取出有用的特征，例如颜色、形状、纹理等。
- 模型训练：利用标注的数据和提取的特征，训练一个深度学习模型。常用的模型包括全连接神经网络（FCN）、支持向量机（SVM）、随机森林等。训练过程中，模型会学习如何根据输入的特征判断鲜花的种类。
- 模型评估与优化：使用一部分未参与训练的数据对模型进行评估，检查模型的准确率、召回率等指标。根据评估结果，对模型进行优化，例如调整网络结构、改变学习率等。
- 模型应用：训练和优化后的模型可以用于实际的鲜花识别任务。对于新的、未标注的图片，模型将自动提取特征并判断其所属的鲜花种类。
通过以上步骤，基于深度学习的鲜花识别方法能够实现较高的识别准确率，有助于花卉产业的智能化管理和运营。
import os
import zipfile
import random
import json
import paddle
import sys
import numpy as np
from PIL import Image
import matplotlib.pyplot as plt
from paddle.io import Datasetimport paddle.fluid as fluid
from visualdl import LogWriter
import re
from paddle.fluid.dygraph import Linear,Conv2D,Pool2D,BatchNorm'''
参数配置
'''
train_parameters = {"input_size": [3, 224, 224], #输入图片的shape"class_dim": -1, #分类数"image_count": -1,"src_path":"D:/save2/data_project/sjwlysdxxdata/flower/flower7595.zip", #原始数据集路径"target_path":"D:/save2/data_project/sjwlysdxxdata/flower/", #要解压的路径"train_list_path": "D:/save2/data_project/sjwlysdxxdata/flower/train.txt", #train.txt路径"eval_list_path": "D:/save2/data_project/sjwlysdxxdata/flower/eval.txt", #eval.txt路径"readme_path": "D:/save2/data_project/sjwlysdxxdata/flower/readme.json", #readme.json路径"label_dict":{}, #标签字典"num_epochs": 100, #训练轮数"train_batch_size": 128, #训练时每个批次的大小"skip_steps": 100,"save_steps": 1000,"learning_strategy": { #优化函数相关的配置"lr": 0.00001 #超参数学习率},"mean_rgb": [127.5, 127.5, 127.5], # 常用图片的三通道均值，通常来说需要先对训练数据做统计，此处仅取中间值"image_enhance_strategy": { # 图像增强相关策略"need_distort": True, # 是否启用图像颜色增强"need_rotate": True, # 是否需要增加随机角度"need_crop": True, # 是否要增加裁剪"need_flip": True, # 是否要增加水平随机翻转"hue_prob": 0.5,"hue_delta": 18,"contrast_prob": 0.5,"contrast_delta": 0.5,"saturation_prob": 0.5,"saturation_delta": 0.5,"brightness_prob": 0.5,"brightness_delta": 0.125},"early_stop": {"sample_frequency": 50,"successive_limit": 3,"good_acc1": 0.92},"rsm_strategy": {"learning_rate": 0.00001,"lr_epochs": [20, 40, 60, 80, 100],"lr_decay": [1, 0.5, 0.25, 0.1, 0.01, 0.002]},"momentum_strategy": {"learning_rate": 0.00001,"lr_epochs": [20, 40, 60, 80, 100],"lr_decay": [1, 0.5, 0.25, 0.1, 0.01, 0.002]},"sgd_strategy": {"learning_rate": 0.00001,"lr_epochs": [20, 40, 60, 80, 100],"lr_decay": [1, 0.5, 0.25, 0.1, 0.01, 0.002]},"adam_strategy": {"learning_rate": 0.000001},"checkpoints": "D:/save2/data_project/sjwlysdxxdata/flower/work/checkpoints" #保存的路径}def unzip_data(src_path,target_path):'''解压原始数据集，将src_path路径下的zip包解压至target_path目录下'''if(not os.path.isdir(target_path + "Chinese Medicine")):z = zipfile.ZipFile(src_path, 'r')z.extractall(path=target_path)z.close()def get_data_list(target_path, train_list_path, eval_list_path):'''生成数据列表'''# 存放所有类别的信息class_detail = []# 获取所有类别保存的文件夹名称data_list_path = target_path + "flowers/"class_dirs = os.listdir(data_list_path)# 总的图像数量all_class_images = 0# 存放类别标签class_label = 0# 存放类别数目class_dim = 0# 存储要写进eval.txt和train.txt中的内容trainer_list = []eval_list = []# 读取每个类别for class_dir in class_dirs:if class_dir != ".DS_Store":class_dim += 1# 每个类别的信息class_detail_list = {}eval_sum = 0trainer_sum = 0# 统计每个类别有多少张图片class_sum = 0# 获取类别路径path = data_list_path + class_dir# 获取所有图片img_paths = os.listdir(path)for img_path in img_paths: # 遍历文件夹下的每个图片if img_path.split(".")[-1] == "jpg":name_path = path + '/' + img_path # 每张图片的路径if class_sum % 8 == 0: # 每8张图片取一个做验证数据eval_sum += 1 # test_sum为测试数据的数目eval_list.append(name_path + "\t%d" % class_label + "\n")else:trainer_sum += 1trainer_list.append(name_path + "\t%d" % class_label + "\n") # trainer_sum测试数据的数目class_sum += 1 # 每类图片的数目all_class_images += 1 # 所有类图片的数目else:continue# 说明的json文件的class_detail数据class_detail_list['class_name'] = class_dir # 类别名称class_detail_list['class_label'] = class_label # 类别标签class_detail_list['class_eval_images'] = eval_sum # 该类数据的测试集数目class_detail_list['class_trainer_images'] = trainer_sum # 该类数据的训练集数目class_detail.append(class_detail_list)# 初始化标签列表train_parameters['label_dict'][str(class_label)] = class_dirclass_label += 1# 初始化分类数train_parameters['class_dim'] = class_dim# 乱序random.shuffle(eval_list)with open(eval_list_path, 'a') as f:for eval_image in eval_list:f.write(eval_image)random.shuffle(trainer_list)with open(train_list_path, 'a') as f2:for train_image in trainer_list:f2.write(train_image)# 说明的json文件信息readjson = {}readjson['all_class_name'] = data_list_path # 文件父目录readjson['all_class_images'] = all_class_imagesreadjson['class_detail'] = class_detailjsons = json.dumps(readjson, sort_keys=True, indent=4, separators=(',', ': '))with open(train_parameters['readme_path'], 'w') as f:f.write(jsons)print('生成数据列表完成！')'''
参数初始化
'''
src_path = train_parameters['src_path']
target_path = train_parameters['target_path']
train_list_path = train_parameters['train_list_path']
eval_list_path = train_parameters['eval_list_path']'''
解压原始数据到指定路径
'''
unzip_data(src_path, target_path)'''
划分训练集与验证集，乱序，生成数据列表
'''
# 每次生成数据列表前，首先清空train.txt和eval.txt
with open(train_list_path, 'w') as f:f.seek(0)f.truncate()
with open(eval_list_path, 'w') as f:f.seek(0)f.truncate()# 生成数据列表
get_data_list(target_path, train_list_path, eval_list_path)class dataset(Dataset):def __init__(self, data_path, mode='train'):"""数据读取器:param data_path: 数据集所在路径:param mode: train or eval"""super().__init__()self.data_path = data_pathself.img_paths = []self.labels = []if mode == 'train':with open(os.path.join(self.data_path, "train.txt"), "r", encoding="utf-8") as f:self.info = f.readlines()for img_info in self.info:img_path, label = img_info.strip().split('\t')self.img_paths.append(img_path)self.labels.append(int(label))else:with open(os.path.join(self.data_path, "eval.txt"), "r", encoding="utf-8") as f:self.info = f.readlines()for img_info in self.info:img_path, label = img_info.strip().split('\t')self.img_paths.append(img_path)self.labels.append(int(label))def __getitem__(self, index):"""获取一组数据:param index: 文件索引号:return:"""# 第一步打开图像文件并获取label值img_path = self.img_paths[index]img = Image.open(img_path)if img.mode != 'RGB':img = img.convert('RGB')img = img.resize((224, 224), Image.BILINEAR)img = np.array(img).astype('float32')img = img.transpose((2, 0, 1)) / 255label = self.labels[index]label = np.array([label], dtype="int64")return img, labeldef print_sample(self, index: int = 0):print("文件名", self.img_paths[index], "\t标签值", self.labels[index])def __len__(self):return len(self.img_paths)#训练数据加载
train_dataset = dataset('D:/save2/data_project/sjwlysdxxdata/flower',mode='train')
train_loader = paddle.io.DataLoader(train_dataset, batch_size=16, shuffle=True)
#测试数据加载
eval_dataset = dataset('D:/save2/data_project/sjwlysdxxdata/flower',mode='eval')
eval_loader = paddle.io.DataLoader(eval_dataset, batch_size = 8, shuffle=False)
train_dataset.print_sample(200)
print(train_dataset.__len__())
eval_dataset.print_sample(0)
print(eval_dataset.__len__())
print(eval_dataset.__getitem__(10)[0].shape)
print(eval_dataset.__getitem__(10)[1].shape)class ConvPool(paddle.nn.Layer):'''卷积+池化'''def __init__(self,num_channels,num_filters,filter_size,pool_size,pool_stride,groups,conv_stride=1,conv_padding=1,):super(ConvPool, self).__init__()for i in range(groups):self.add_sublayer( # 添加子层实例'bb_%d' % i,paddle.nn.Conv2D( # layerin_channels=num_channels, # 通道数out_channels=num_filters, # 卷积核个数kernel_size=filter_size, # 卷积核大小stride=conv_stride, # 步长padding=conv_padding, # padding))self.add_sublayer('relu%d' % i,paddle.nn.ReLU())num_channels = num_filtersself.add_sublayer('Maxpool',paddle.nn.MaxPool2D(kernel_size=pool_size, # 池化核大小stride=pool_stride # 池化步长))def forward(self, inputs):x = inputsfor prefix, sub_layer in self.named_children():# print(prefix,sub_layer)x = sub_layer(x)return xclass VGGNet(paddle.nn.Layer):def __init__(self):super(VGGNet, self).__init__()self.convpool01 = ConvPool(3, 64, 3, 2, 2, 2) # 3:通道数，64：卷积核个数，3:卷积核大小，2:池化核大小，2:池化步长，2:连续卷积个数self.convpool02 = ConvPool(64, 128, 3, 2, 2, 2)self.convpool03 = ConvPool(128, 256, 3, 2, 2, 3)self.convpool04 = ConvPool(256, 512, 3, 2, 2, 3)self.convpool05 = ConvPool(512, 512, 3, 2, 2, 3)self.pool_5_shape = 512 * 7 * 7self.fc01 = paddle.nn.Linear(self.pool_5_shape, 4096)self.fc02 = paddle.nn.Linear(4096, 4096)self.fc03 = paddle.nn.Linear(4096, train_parameters['class_dim'])def forward(self, inputs, label=None):# print('input_shape:', inputs.shape) #[8, 3, 224, 224]"""前向计算"""out = self.convpool01(inputs)# print('convpool01_shape:', out.shape) #[8, 64, 112, 112]out = self.convpool02(out)# print('convpool02_shape:', out.shape) #[8, 128, 56, 56]out = self.convpool03(out)# print('convpool03_shape:', out.shape) #[8, 256, 28, 28]out = self.convpool04(out)# print('convpool04_shape:', out.shape) #[8, 512, 14, 14]out = self.convpool05(out)# print('convpool05_shape:', out.shape) #[8, 512, 7, 7]out = paddle.reshape(out, shape=[-1, 512 * 7 * 7])out = self.fc01(out)out = self.fc02(out)out = self.fc03(out)if label is not None:acc = paddle.metric.accuracy(input=out, label=label)return out, accelse:return outdef draw_process(title,color,iters,data,label):plt.title(title, fontsize=24)plt.xlabel("iter", fontsize=20)plt.ylabel(label, fontsize=20)plt.plot(iters, data,color=color,label=label)plt.legend()plt.grid()plt.show()print(train_parameters['class_dim'])
print(train_parameters['label_dict'])model = VGGNet()
model.train()
cross_entropy = paddle.nn.CrossEntropyLoss()
optimizer = paddle.optimizer.Adam(learning_rate=train_parameters['learning_strategy']['lr'],parameters=model.parameters())steps = 0
Iters, total_loss, total_acc = [], [], []for epo in range(train_parameters['num_epochs']):for _, data in enumerate(train_loader()):steps += 1x_data = data[0]y_data = data[1]predicts, acc = model(x_data, y_data)loss = cross_entropy(predicts, y_data)loss.backward()optimizer.step()optimizer.clear_grad()if steps % train_parameters["skip_steps"] == 0:Iters.append(steps)total_loss.append(loss.numpy()[0])total_acc.append(acc.numpy()[0])#打印中间过程print('epo: {}, step: {}, loss is: {}, acc is: {}'\.format(epo, steps, loss.numpy(), acc.numpy()))#保存模型参数if steps % train_parameters["save_steps"] == 0:save_path = train_parameters["checkpoints"]+"/"+"save_dir_" + str(steps) + '.pdparams'print('save model to: ' + save_path)paddle.save(model.state_dict(),save_path)
paddle.save(model.state_dict(),train_parameters["checkpoints"]+"/"+"save_dir_final.pdparams")
draw_process("trainning loss","red",Iters,total_loss,"trainning loss")
draw_process("trainning acc","green",Iters,total_acc,"trainning acc")'''
模型评估
'''
model__state_dict = paddle.load('work/checkpoints/save_dir_final.pdparams')
model_eval = VGGNet()
model_eval.set_state_dict(model__state_dict)
model_eval.eval()
accs = []for _, data in enumerate(eval_loader()):x_data = data[0]y_data = data[1]predicts = model_eval(x_data)acc = paddle.metric.accuracy(predicts, y_data)accs.append(acc.numpy()[0])
print('模型在验证集上的准确率为：',np.mean(accs))def load_image(img_path):'''预测图片预处理'''img = Image.open(img_path)if img.mode != 'RGB':img = img.convert('RGB')img = img.resize((224, 224), Image.BILINEAR)img = np.array(img).astype('float32')img = img.transpose((2, 0, 1)) / 255 # HWC to CHW 及归一化return imginfer_dst_path = 'data/flowers/rose/'label_dic = train_parameters['label_dict']model__state_dict = paddle.load('work/checkpoints/save_dir_final.pdparams')
model_predict = VGGNet()
model_predict.set_state_dict(model__state_dict)
model_predict.eval()
infer_imgs_path = os.listdir(infer_dst_path)
#print(infer_imgs_path)
for infer_img_path in infer_imgs_path[:10]:infer_img = load_image(infer_dst_path+infer_img_path)infer_img = infer_img[np.newaxis,:, : ,:] #reshape(-1,3,224,224)infer_img = paddle.to_tensor(infer_img)result = model_predict(infer_img)lab = np.argmax(result.numpy())print("rose样本: {},被预测为:{}".format(infer_img_path,label_dic[str(lab)]))
