当前位置： 首页 > news >正文

网站建设 中企动力西安,北京网站建设q.479185700強,域名申请到网站上传全过程,wordpress小程序搭建【PyTorch】基于YOLO的多目标检测项目#xff08;一#xff09; 【PyTorch】基于YOLO的多目标检测项目#xff08;二#xff09; 目标检测是对图像中的现有目标进行定位和分类的过程。识别的对象在图像中显示有边界框。一般的目标检测方法有两种#xff1a;基于区域提议的… 【PyTorch】基于YOLO的多目标检测项目一 【PyTorch】基于YOLO的多目标检测项目二 目标检测是对图像中的现有目标进行定位和分类的过程。识别的对象在图像中显示有边界框。一般的目标检测方法有两种基于区域提议的和基于回归/分类的。这里使用一种基于回归/分类的方法称为YOLO。 目录 准备COCO数据集 创建自定义数据集 转换数据 定义数据加载器 准备COCO数据集 COCO是一个大规模的对象检测分割和字幕数据集。它包含80个对象类别用于对象检测。 下载以下GitHub存储库 https://github.com/pjreddie/darknethttps://github.com/pjreddie/darknet 创建一个名为config的文件夹将darknet/cfg/coco.data、darknet/cfg/yolov3.cfg文件复制到config文件夹中。 创建一个名为data的文件夹从以下链接获取coco.names文件并将其放入data文件夹coco.names文件包含COCO数据集中80个对象类别的列表。 darknet/data/coco.names at master · pjreddie/darknet · GitHubConvolutional Neural Networks. Contribute to pjreddie/darknet development by creating an account on GitHub.https://github.com/pjreddie/darknet/blob/master/data/coco.names将darknet/scripts/get_coco_dataset.sh文件复制到data文件夹中并复制get_coco_cocoet.sh到data文件夹。接下来打开一个终端并执行get_coco_cocoet.sh该脚本将把完整的COCO数据集下载到名为coco的子文件夹中。也可通过以下链接下载coco数据集。 COCO2014_数据集-飞桨AI Studio星河社区 (baidu.com)https://aistudio.baidu.com/datasetdetail/165195 在images文件夹中有两个名为train 2014和val 2014的文件夹分别包含82783和40504个图像。在labels文件夹中有两个名为train 2014和val 2014的标签分别包含82081和40137文本文件。这些文本文件包含图像中对象的边界框坐标。此外trainvalno5k.txt文件是一个包含117264张图像的列表这些图像将用于训练模型。此列表是train2014和val2014中图像的组合5000个图像除外。5k.txt文件包含将用于验证的5000个图像的列表。 创建自定义数据集 完成数据集下载后使用PyTorch的Dataset和Dataloader类创建训练和验证数据集和数据加载器。 from torch.utils.data import Dataset from PIL import Image import torchvision.transforms.functional as TF import os import numpy as npimport torch device torch.device(cuda:0 if torch.cuda.is_available() else cpu) print(torch.version) #定义CocoDataset类并展示来自训练和验证数据集的一些示例图像 class CocoDataset(Dataset):def init(self, path2listFile, transformNone, trans_paramsNone):with open(path2listFile, r) as file:self.path2imgs file.readlines()self.path2labels [path.replace(images, labels).replace(.png, .txt).replace(.jpg, .txt)for path in self.path2imgs]self.trans_params trans_paramsself.transform transformdef len(self):return len(self.path2imgs)def getitem(self, index):path2img self.path2imgs[index % len(self.path2imgs)].rstrip()img Image.open(path2img).convert(RGB)path2label self.path2labels[index % len(self.path2imgs)].rstrip()labels Noneif os.path.exists(path2label):labels np.loadtxt(path2label).reshape(-1, 5)if self.transform:img, labels self.transform(img, labels, self.trans_params)return img, labels, path2img
root_data./data/coco path2trainListos.path.join(root_data, trainvalno5k.txt)coco_train CocoDataset(path2trainList) print(len(coco_train)) # 从coco_train中获取图像、标签和图像路径 img, labels, path2img coco_train[1] print(image size:, img.size, type(img)) print(labels shape:, labels.shape, type(labels)) print(labels \n, labels) path2valListos.path.join(root_data, 5k.txt) coco_val CocoDataset(path2valList, transformNone, trans_paramsNone) print(len(coco_val)) img, labels, path2img coco_val[7] print(image size:, img.size, type(img)) print(labels shape:, labels.shape, type(labels)) print(labels \n, labels) import matplotlib.pylab as plt import numpy as np from PIL import Image, ImageDraw, ImageFont from torchvision.transforms.functional import to_pil_image import random %matplotlib inline path2cocoNames./data/coco.names fp open(path2cocoNames, r) coco_names fp.read().split(\n)[:-1] print(number of classese:, len(coco_names)) print(coco_names) def rescale_bbox(bb,W,H):x,y,w,hbbreturn [x*W, y*H, w*W, h*H] COLORS np.random.randint(0, 255, size(80, 3),dtypeuint8)

fnt ImageFont.truetype(Pillow/Tests/fonts/FreeMono.ttf, 16)

fnt ImageFont.truetype(arial.ttf, 16) def show_img_bbox(img,targets):if torch.is_tensor(img):imgto_pil_image(img)if torch.is_tensor(targets):targetstargets.numpy()[:,1:]W, Himg.sizedraw ImageDraw.Draw(img)for tg in targets:id_int(tg[0])bboxtg[1:]bboxrescalebbox(bbox,W,H)xc,yc,w,hbboxcolor [int© for c in COLORS[id]]namecoconames[id]draw.rectangle(((xc-w/2, yc-h/2), (xcw/2, ych/2)),outlinetuple(color),width3)draw.text((xc-w/2,yc-h/2),name, fontfnt, fill(255,255,255,0))plt.imshow(np.array(img))
np.random.seed(1) rnd_indnp.random.randint(len(coco_train)) img, labels, path2img coco_train[rnd_ind] print(img.size, labels.shape)plt.rcParamsfigure.figsize show_img_bbox(img,labels) np.random.seed(1) rnd_indnp.random.randint(len(coco_val)) img, labels, path2img coco_val[rnd_ind] print(img.size, labels.shape)plt.rcParamsfigure.figsize show_img_bbox(img,labels) 转换数据 定义一个转换函数和传递给CocoDataset类的参数 def pad_to_square(img, boxes, pad_value0, normalized_labelsTrue):w, h img.sizew_factor, h_factor (w,h) if normalized_labels else (1, 1)dim_diff np.abs(h - w)pad1 dim_diff // 2pad2 dim_diff - pad1if hw:left, top, right, bottom 0, pad1, 0, pad2else:left, top, right, bottom pad1, 0, pad2, 0padding (left, top, right, bottom)img_padded TF.pad(img, paddingpadding, fillpad_value)w_padded, h_padded img_padded.sizex1 w_factor * (boxes[:, 1] - boxes[:, 3] / 2)y1 h_factor * (boxes[:, 2] - boxes[:, 4] / 2)x2 w_factor * (boxes[:, 1] boxes[:, 3] / 2)y2 h_factor * (boxes[:, 2] boxes[:, 4] / 2) x1 padding[0] # 左y1 padding[1] # 上x2 padding[2] # 右y2 padding[3] # 下boxes:, 1 / w_paddedboxes:, 2 / h_paddedboxes[:, 3] * w_factor / w_paddedboxes[:, 4] * h_factor / h_paddedreturn img_padded, boxes
def hflip(image, labels):image TF.hflip(image)labels[:, 1] 1.0 - labels[:, 1]return image, labelsdef transformer(image, labels, params):if params[pad2square] is True:image,labels pad_to_square(image, labels)image TF.resize(image,params[target_size])if random.random() params[p_hflip]:image,labelshflip(image,labels)imageTF.to_tensor(image)targets torch.zeros((len(labels), 6))targets[:, 1:] torch.from_numpy(labels)return image, targets trans_params_train{target_size : (416, 416),pad2square: True,p_hflip : 1.0,normalized_labels: True, } coco_trainCocoDataset(path2trainList,transformtransformer,trans_paramstrans_params_train)np.random.seed(100) rnd_indnp.random.randint(len(coco_train)) img, targets, path2img coco_train[rnd_ind] print(image shape:, img.shape) print(labels shape:, targets.shape) plt.rcParamsfigure.figsize COLORS np.random.randint(0, 255, size(80, 3),dtypeuint8) show_img_bbox(img,targets) 通过传递 transformer 函数来定义 CocoDataset 的一个对象来验证数据  trans_params_val{target_size : (416, 416),pad2square: True,p_hflip : 0.0,normalized_labels: True, } coco_val CocoDataset(path2valList,transformtransformer,trans_paramstrans_params_val)np.random.seed(55) rnd_indnp.random.randint(len(coco_val)) img, targets, path2img coco_val[rnd_ind] print(image shape:, img.shape) print(labels shape:, targets.shape) plt.rcParamsfigure.figsize COLORS np.random.randint(0, 255, size(80, 3),dtypeuint8) show_img_bbox(img,targets) 定义数据加载器 定义两个用于训练和验证数据集的数据加载器从coco_train和coco_val中获取小批量数据。 from torch.utils.data import DataLoaderbatch_size8 def collate_fn(batch):imgs, targets, paths list(zip(*batch))targets [boxes for boxes in targets if boxes is not None]for b_i, boxes in enumerate(targets):boxes[:, 0] b_itargets torch.cat(targets, 0)imgs torch.stack([img for img in imgs])return imgs, targets, pathstrain_dl DataLoader(coco_train,batch_sizebatch_size,shuffleTrue,num_workers0,pin_memoryTrue,collate_fncollate_fn,)torch.manual_seed(0) for imgs_batch,tg_batch,path_batch in train_dl:break print(imgs_batch.shape) print(tg_batch.shape,tg_batch.dtype) val_dl DataLoader(coco_val,batch_sizebatch_size,shuffleFalse,num_workers0,pin_memoryTrue,collate_fncollate_fn,)torch.manual_seed(0) for imgs_batch,tg_batch,path_batch in val_dl:break print(imgs_batch.shape) print(tg_batch.shape,tg_batch.dtype)