YOLOV5源码解读系列文章目录

数据集加载和增强
loss计算

前言

此篇为yolov5 3.1 版本，官方地址[https://github.com/ultralytics/yolov5]
看源代码之前有必要先大致了解实现原理和流程，强推这篇文章https://blog.csdn.net/nan355655600/article/details/107852353(https://github.com/amdegroot/ssd.pytorch)

数据加载器由utils/datasets.py文件中的create_dataloader函数创建，其中主要有两个类构成LoadImagesAndLabels：数据集的加载和增强都由这个类实现 InfiniteDataLoader：对DataLoader进行封装，就是为了能够永久持续的采样数据，详细原因这里可以看官方说明[https://github.com/ultralytics/yolov5/pull/876](https://github.com/ultralytics/yolov5/pull/876)

持续采样InfiniteDataLoader

class InfiniteDataLoader(torch.utils.data.dataloader.DataLoader):""" Dataloader that reuses workersUses same syntax as vanilla DataLoader""""""这块对DataLoader进行封装，就是为了能够永久持续的采样数据"""def __init__(self, *args, **kwargs):super().__init__(*args, **kwargs)object.__setattr__(self, 'batch_sampler', _RepeatSampler(self.batch_sampler))self.iterator = super().__iter__()def __len__(self):return len(self.batch_sampler.sampler)def __iter__(self):for i in range(len(self)):yield next(self.iterator)class _RepeatSampler(object):""" Sampler that repeats forever永久持续的采样Args:sampler (Sampler)"""def __init__(self, sampler):self.sampler = samplerdef __iter__(self):while True:yield from iter(self.sampler)

数据加载

class LoadImagesAndLabels(Dataset):  # for training/testingdef __init__(self, path, img_size=640, batch_size=16, augment=False, hyp=None, rect=False, image_weights=False,cache_images=False, single_cls=False, stride=32, pad=0.0, rank=-1):"""path:    数据集路径img_size:    图片大小batch_size:  批次大小augment: 是否数据增强hyp: 超参数的yaml文件rect: 矩形训练，就是对图片填充灰边（只在高或宽的一边填充）image_weights: 图像采样的权重cache_images: 图片是否缓存，用于加速训练single_cls: 是否是一个类别stride: 模型步幅， 图像大小/网络下采样之后的输出大小pad: 填充宽度rank: 当前进程编号"""self.img_size = img_sizeself.augment = augmentself.hyp = hypself.image_weights = image_weightsself.rect = False if image_weights else rect# mosaic 将4张图片融合在一张图片里，进行训练self.mosaic = self.augment and not self.rect  # load 4 images at a time into a mosaic (only during training)self.mosaic_border = [-img_size // 2, -img_size // 2]self.stride = stride"""首先读取图像路径，转换合适的格式，根据图像路径，替换其中的images和图片后缀，转换成label路径读取coco128/labels/train.cache文件,没有则创建，cache存储字典{图片路径:label路径，图片大小}"""def img2label_paths(img_paths):# Define label paths as a function of image paths"""img_paths现在存储了所有的图片路径，只需将路径中的images换成labels，图片后缀改为.txt就得到标注文件的路径"""sa, sb = os.sep + 'images' + os.sep, os.sep + 'labels' + os.sep  # /images/, /labels/ substringsreturn [x.replace(sa, sb, 1).replace(os.path.splitext(x)[-1], '.txt') for x in img_paths]# 读取图像路径，转换成合适的格式try:f = []  # image filesfor p in path if isinstance(path, list) else [path]:p = str(Path(p))  # os-agnosticparent = str(Path(p).parent) + os.sep   #上级目录  ../coco128/imagesif os.path.isfile(p):  # filewith open(p, 'r') as t:t = t.read().splitlines()f += [x.replace('./', parent) if x.startswith('./') else x for x in t]  # local to global pathelif os.path.isdir(p):  # folderf += glob.iglob(p + os.sep + '*.*')     # 读取images下的所有文件不包含目录else:raise Exception('%s does not exist' % p)# 将图片的路径改为适合本地系统的格式（windows是'\\', linux是'/')，图片后缀名在img_formats里的就改为小写self.img_files = sorted([x.replace('/', os.sep) for x in f if os.path.splitext(x)[-1].lower() in img_formats])assert len(self.img_files) > 0, 'No images found'except Exception as e:raise Exception('Error loading data from %s: %s\nSee %s' % (path, e, help_url))# Check cacheself.label_files = img2label_paths(self.img_files)  # labels 图片路径到label路径的转换cache_path = str(Path(self.label_files[0]).parent) + '.cache'  # cached labels"""读取labels下的.cache文件， 没有则创建， cache里的关键字'hash'是图片+label的文件字节大小之和"""if os.path.isfile(cache_path):cache = torch.load(cache_path)  # load# 如果cache存储的hash与当前的label+图片大小对应不上,则重新创建.cache文件if cache['hash'] != get_hash(self.label_files + self.img_files):  # dataset changedcache = self.cache_labels(cache_path)  # re-cacheelse:cache = self.cache_labels(cache_path)  # cache# Read cachecache.pop('hash')  # remove hashlabels, shapes = zip(*cache.values())self.labels = list(labels)      # labelself.shapes = np.array(shapes, dtype=np.float64)    # 图片大小self.img_files = list(cache.keys())  # update   图片路径self.label_files = img2label_paths(cache.keys())  # update  更新labels路径，因为可能有一部分图片或label损坏"""根据图片数量划分每批的图片数量"""n = len(shapes)  # number of images     图片数量bi = np.floor(np.arange(n) / batch_size).astype(np.int)  # batch index  划分批次nb = bi[-1] + 1  # number of batches    批次数量self.batch = bi  # batch index of imageself.n = n# Rectangular Training  矩形训练"""先求的图像的宽高比，然后对较长的边缩放到stride的倍数，在按照宽高比对短的一边缩放，进行少量的填充也达到stride的最小倍数"""if self.rect:# Sort by aspect ratios = self.shapes  # whar = s[:, 1] / s[:, 0]  # aspect ratio  高宽比irect = ar.argsort()    # 按着高宽比从小到大排序# 重新排序图片，label路径，真实框, shapes, 宽高比的顺序self.img_files = [self.img_files[i] for i in irect]self.label_files = [self.label_files[i] for i in irect]self.labels = [self.labels[i] for i in irect]self.shapes = s[irect]  # whar = ar[irect]# Set training image shapesshapes = [[1, 1]] * nb  # [[h/w, 1], [1, w/h]....]for i in range(nb):ari = ar[bi == i]   # 分批次选择mini, maxi = ari.min(), ari.max()# 下面操作就是为了保证shapes存储的值始终小于1,即只对较短的一边进行操作if maxi < 1:    # 高小于宽的时候shapes[i] = [maxi, 1]elif mini > 1:  # 高大于宽的时候shapes[i] = [1, 1 / mini]# 下边的操作就是为了得到以stride为整数倍的图像大小（较短的一边）注意：只在测试时才会用到self.batch_shapes = np.ceil(np.array(shapes) * img_size / stride + pad).astype(np.int) * stride# Check labels 检查标签是否合法create_datasubset, extract_bounding_boxes, labels_loaded = False, False, False  # 目前这些操作还不支持# 消失的，存在的，空的，小型数据集的，重复的标签数量nm, nf, ne, ns, nd = 0, 0, 0, 0, 0  # number missing, found, empty, datasubset, duplicatepbar = enumerate(self.label_files)if rank in [-1, 0]:pbar = tqdm(pbar)for i, file in pbar:l = self.labels[i]  # labelif l is not None and l.shape[0]:assert l.shape[1] == 5, '> 5 label columns: %s' % file  # 类别+4个坐标assert (l >= 0).all(), 'negative labels: %s' % file     # 不能出现负数assert (l[:, 1:] <= 1).all(), 'non-normalized or out of bounds coordinate labels: %s' % file    # 坐标不能大于1if np.unique(l, axis=0).shape[0] < l.shape[0]:  # duplicate rows  出现重复的标签nd += 1  # print('WARNING: duplicate rows in %s' % self.label_files[i])  # duplicate rowsif single_cls:  # 如果设置单一类别，则将label所有类别设为0l[:, 0] = 0  # force dataset into single-class modeself.labels[i] = lnf += 1  # file found# Create subdataset (a smaller dataset)if create_datasubset and ns < 1E4:if ns == 0:create_folder(path='./datasubset')os.makedirs('./datasubset/images')exclude_classes = 43if exclude_classes not in l[:, 0]:ns += 1# shutil.copy(src=self.img_files[i], dst='./datasubset/images/')  # copy imagewith open('./datasubset/images.txt', 'a') as f:f.write(self.img_files[i] + '\n')# Extract object detection boxes for a second stage classifierif extract_bounding_boxes:p = Path(self.img_files[i])img = cv2.imread(str(p))h, w = img.shape[:2]for j, x in enumerate(l):f = '%s%sclassifier%s%g_%g_%s' % (p.parent.parent, os.sep, os.sep, x[0], j, p.name)if not os.path.exists(Path(f).parent):os.makedirs(Path(f).parent)  # make new output folderb = x[1:] * [w, h, w, h]  # boxb[2:] = b[2:].max()  # rectangle to squareb[2:] = b[2:] * 1.3 + 30  # padb = xywh2xyxy(b.reshape(-1, 4)).ravel().astype(np.int)b[[0, 2]] = np.clip(b[[0, 2]], 0, w)  # clip boxes outside of imageb[[1, 3]] = np.clip(b[[1, 3]], 0, h)assert cv2.imwrite(f, img[b[1]:b[3], b[0]:b[2]]), 'Failure extracting classifier boxes'else:ne += 1  # print('empty labels for image %s' % self.img_files[i])  # file empty# os.system("rm '%s' '%s'" % (self.img_files[i], self.label_files[i]))  # removeif rank in [-1, 0]:pbar.desc = 'Scanning labels %s (%g found, %g missing, %g empty, %g duplicate, for %g images)' % (cache_path, nf, nm, ne, nd, n)if nf == 0:s = 'WARNING: No labels found in %s. See %s' % (os.path.dirname(file) + os.sep, help_url)print(s)assert not augment, '%s. Can not train without labels.' % s# Cache images into memory for faster training (WARNING: large datasets may exceed system RAM)# 缓存图像到内存中，为了快速训练, 通过调用8个线程，读取图像并进行resize处理，保存在self.imgs变量中self.imgs = [None] * nif cache_images:gb = 0  # Gigabytes of cached imagesself.img_hw0, self.img_hw = [None] * n, [None] * n  # 原始图片大小,resize之后的图片大小results = ThreadPool(8).imap(lambda x: load_image(*x), zip(repeat(self), range(n)))  # 8 threadspbar = tqdm(enumerate(results), total=n)for i, x in pbar:self.imgs[i], self.img_hw0[i], self.img_hw[i] = x  # img, hw_original, hw_resized = load_image(self, i)gb += self.imgs[i].nbytespbar.desc = 'Caching images (%.1fGB)' % (gb / 1E9)def cache_labels(self, path='labels.cache'):# Cache dataset labels, check images and read shapes"""检测image和label有没有损坏"""x = {}  # dictpbar = tqdm(zip(self.img_files, self.label_files), desc='Scanning images', total=len(self.img_files))   # 进度条for (img, label) in pbar:try:l = []im = Image.open(img)im.verify()  # PIL verify   判断图像是否损坏shape = exif_size(im)  # image sizeassert (shape[0] > 9) & (shape[1] > 9), 'image size <10 pixels'if os.path.isfile(label):with open(label, 'r') as f:l = np.array([x.split() for x in f.read().splitlines()], dtype=np.float32)  # labelsif len(l) == 0:l = np.zeros((0, 5), dtype=np.float32)x[img] = [l, shape]except Exception as e:print('WARNING: Ignoring corrupted image and/or label %s: %s' % (img, e))x['hash'] = get_hash(self.label_files + self.img_files)     # 图像+label的文件字节大小torch.save(x, path)  # save for next timereturn xdef __len__(self):return len(self.img_files)# def __iter__(self):#     self.count = -1#     print('ran dataset iter')#     #self.shuffled_vector = np.random.permutation(self.nF) if self.augment else np.arange(self.nF)#     return selfdef __getitem__(self, index):if self.image_weights:index = self.indices[index]hyp = self.hypmosaic = self.mosaic and random.random() < hyp['mosaic']if mosaic:# Load mosaicimg, labels = load_mosaic(self, index)shapes = None# MixUp https://arxiv.org/pdf/1710.09412.pdf# 将两幅图片混合在一起（每幅图片包含4张小图）if random.random() < hyp['mixup']:img2, labels2 = load_mosaic(self, random.randint(0, len(self.labels) - 1))r = np.random.beta(8.0, 8.0)  # mixup ratio, alpha=beta=8.0img = (img * r + img2 * (1 - r)).astype(np.uint8)labels = np.concatenate((labels, labels2), 0)else:# Load imageimg, (h0, w0), (h, w) = load_image(self, index)# Letterboxshape = self.batch_shapes[self.batch[index]] if self.rect else self.img_size  # final letterboxed shapeimg, ratio, pad = letterbox(img, shape, auto=False, scaleup=self.augment)shapes = (h0, w0), ((h / h0, w / w0), pad)  # for COCO mAP rescaling# Load labelslabels = []x = self.labels[index]if x.size > 0:# Normalized xywh to pixel xyxy format# 将标签格式[centerx, centery, w, h]转换为[xim, ymin, xmax, ymax],# 并调整为未归一化的格式(图片上真实坐标), 坐标平移调整labels = x.copy()labels[:, 1] = ratio[0] * w * (x[:, 1] - x[:, 3] / 2) + pad[0]  # pad widthlabels[:, 2] = ratio[1] * h * (x[:, 2] - x[:, 4] / 2) + pad[1]  # pad heightlabels[:, 3] = ratio[0] * w * (x[:, 1] + x[:, 3] / 2) + pad[0]labels[:, 4] = ratio[1] * h * (x[:, 2] + x[:, 4] / 2) + pad[1]if self.augment:# Augment imagespaceif not mosaic:img, labels = random_perspective(img, labels,degrees=hyp['degrees'],translate=hyp['translate'],scale=hyp['scale'],shear=hyp['shear'],perspective=hyp['perspective'])# Augment colorspaceaugment_hsv(img, hgain=hyp['hsv_h'], sgain=hyp['hsv_s'], vgain=hyp['hsv_v'])# Apply cutouts# if random.random() < 0.9:#     labels = cutout(img, labels)nL = len(labels)  # number of labelsif nL:labels[:, 1:5] = xyxy2xywh(labels[:, 1:5])  # convert xyxy to xywhlabels[:, [2, 4]] /= img.shape[0]  # normalized height 0-1labels[:, [1, 3]] /= img.shape[1]  # normalized width 0-1if self.augment:# flip up-down  垂直翻转if random.random() < hyp['flipud']:img = np.flipud(img)if nL:labels[:, 2] = 1 - labels[:, 2]# flip left-right  水平翻转if random.random() < hyp['fliplr']:img = np.fliplr(img)if nL:labels[:, 1] = 1 - labels[:, 1]labels_out = torch.zeros((nL, 6))   # [num_labels, batch_index, class_id, x, y, w, h]if nL:labels_out[:, 1:] = torch.from_numpy(labels)# Convertimg = img[:, :, ::-1].transpose(2, 0, 1)  # BGR to RGB, to 3x416x416img = np.ascontiguousarray(img)return torch.from_numpy(img), labels_out, self.img_files[index], shapes@staticmethoddef collate_fn(batch):"""pytorch的DataLoader打包一个batch的数据集时要经过此函数进行打包通过重写此函数实现标签与图片对应的划分，一个batch中哪些标签属于哪一张图片"""img, label, path, shapes = zip(*batch)  # transposedfor i, l in enumerate(label):l[:, 0] = i  # add target image index for build_targets()return torch.stack(img, 0), torch.cat(label, 0), path, shapes

masico数据增强

def load_mosaic(self, index):# loads images in a mosaiclabels4 = []s = self.img_size# 随机选取mosaic的中心点yc, xc = [int(random.uniform(-x, 2 * s + x)) for x in self.mosaic_border]  # mosaic center x, y# 随机添加剩余3张图像序列indices = [index] + [random.randint(0, len(self.labels) - 1) for _ in range(3)]  # 3 additional image indicesfor i, index in enumerate(indices):# Load imageimg, _, (h, w) = load_image(self, index)# place img in img4  融合4张图片if i == 0:  # top left  左上角图片  114:代表灰色img4 = np.full((s * 2, s * 2, img.shape[2]), 114, dtype=np.uint8)  # base image with 4 tiles# 当前图像在一张大图上的位置x1a, y1a, x2a, y2a = max(xc - w, 0), max(yc - h, 0), xc, yc  # xmin, ymin, xmax, ymax (large image)# 选取当前图像的位置x1b, y1b, x2b, y2b = w - (x2a - x1a), h - (y2a - y1a), w, h  # xmin, ymin, xmax, ymax (small image)elif i == 1:  # top right   右上角图片x1a, y1a, x2a, y2a = xc, max(yc - h, 0), min(xc + w, s * 2), ycx1b, y1b, x2b, y2b = 0, h - (y2a - y1a), min(w, x2a - x1a), helif i == 2:  # bottom left  左下角图片x1a, y1a, x2a, y2a = max(xc - w, 0), yc, xc, min(s * 2, yc + h)x1b, y1b, x2b, y2b = w - (x2a - x1a), 0, w, min(y2a - y1a, h)elif i == 3:  # bottom right   右下角图片x1a, y1a, x2a, y2a = xc, yc, min(xc + w, s * 2), min(s * 2, yc + h)x1b, y1b, x2b, y2b = 0, 0, min(w, x2a - x1a), min(y2a - y1a, h)# 将当前图像的候选区域赋值给大图上设置好的区域img4[y1a:y2a, x1a:x2a] = img[y1b:y2b, x1b:x2b]  # img4[ymin:ymax, xmin:xmax]padw = x1a - x1b    # 截取之后的图像相对原始图像的偏移量padh = y1a - y1b# Labelsx = self.labels[index]labels = x.copy()# label+上面计算好的偏移量if x.size > 0:  # Normalized xywh to pixel xyxy formatlabels[:, 1] = w * (x[:, 1] - x[:, 3] / 2) + padwlabels[:, 2] = h * (x[:, 2] - x[:, 4] / 2) + padhlabels[:, 3] = w * (x[:, 1] + x[:, 3] / 2) + padwlabels[:, 4] = h * (x[:, 2] + x[:, 4] / 2) + padhlabels4.append(labels)# Concat/clip labelsif len(labels4):labels4 = np.concatenate(labels4, 0)# 对大图裁剪，对超出图像边界的值赋予0或img_sizenp.clip(labels4[:, 1:], 0, 2 * s, out=labels4[:, 1:])  # use with random_perspective# img4, labels4 = replicate(img4, labels4)  # replicate# Augment  对图像和标签进行平移，旋转，透视等等处理img4, labels4 = random_perspective(img4, labels4,degrees=self.hyp['degrees'],translate=self.hyp['translate'],scale=self.hyp['scale'],shear=self.hyp['shear'],perspective=self.hyp['perspective'],border=self.mosaic_border)  # border to removereturn img4, labels4

仿射变换

def random_perspective(img, targets=(), degrees=10, translate=.1, scale=.1, shear=10, perspective=0.0, border=(0, 0)):# torchvision.transforms.RandomAffine(degrees=(-10, 10), translate=(.1, .1), scale=(.9, 1.1), shear=(-10, 10))# targets = [cls, xyxy]height = img.shape[0] + border[0] * 2  # shape(h,w,c)width = img.shape[1] + border[1] * 2# Center  让图片先放在正中央的位置，再进行缩放等处理C = np.eye(3)C[0, 2] = -img.shape[1] / 2  # x translation (pixels)C[1, 2] = -img.shape[0] / 2  # y translation (pixels)# Perspective  透视P = np.eye(3)P[2, 0] = random.uniform(-perspective, perspective)  # x perspective (about y)P[2, 1] = random.uniform(-perspective, perspective)  # y perspective (about x)# Rotation and Scale  旋转和缩放R = np.eye(3)a = random.uniform(-degrees, degrees)# a += random.choice([-180, -90, 0, 90])  # add 90deg rotations to small rotationss = random.uniform(1 - scale, 1 + scale)# s = 2 ** random.uniform(-scale, scale)R[:2] = cv2.getRotationMatrix2D(angle=a, center=(0, 0), scale=s)# Shear 错切S = np.eye(3)S[0, 1] = math.tan(random.uniform(-shear, shear) * math.pi / 180)  # x shear (deg)S[1, 0] = math.tan(random.uniform(-shear, shear) * math.pi / 180)  # y shear (deg)# Translation  平移T = np.eye(3)T[0, 2] = random.uniform(0.5 - translate, 0.5 + translate) * width  # x translation (pixels)T[1, 2] = random.uniform(0.5 - translate, 0.5 + translate) * height  # y translation (pixels)# Combined rotation matrix# @:线性代数的矩阵乘法操作  M:变换矩阵M = T @ S @ R @ P @ C  # order of operations (right to left) is IMPORTANTif (border[0] != 0) or (border[1] != 0) or (M != np.eye(3)).any():  # image changedif perspective:img = cv2.warpPerspective(img, M, dsize=(width, height), borderValue=(114, 114, 114))else:  # affine     仿射变换img = cv2.warpAffine(img, M[:2], dsize=(width, height), borderValue=(114, 114, 114))# Visualize# import matplotlib.pyplot as plt# ax = plt.subplots(1, 2, figsize=(12, 6))[1].ravel()# ax[0].imshow(img[:, :, ::-1])  # base# ax[1].imshow(img2[:, :, ::-1])  # warped# Transform label coordinates# 相应的label也要转换n = len(targets)    # label数量   [num_labels, 5]if n:# warp pointsxy = np.ones((n * 4, 3))# targets 坐标形式是[xmin, ymin, xmax, ymax]  下边这句话就是提取真实框的四个点xy[:, :2] = targets[:, [1, 2, 3, 4, 1, 4, 3, 2]].reshape(n * 4, 2)  # x1y1, x2y2, x1y2, x2y1# 注意:下面的T是矩阵转置,而不是上边用于仿射变换的矩阵Txy = xy @ M.T  # transformif perspective:xy = (xy[:, :2] / xy[:, 2:3]).reshape(n, 8)  # rescaleelse:  # affinexy = xy[:, :2].reshape(n, 8)# create new boxes  得到新的真实框x = xy[:, [0, 2, 4, 6]]y = xy[:, [1, 3, 5, 7]]xy = np.concatenate((x.min(1), y.min(1), x.max(1), y.max(1))).reshape(4, n).T# # apply angle-based reduction of bounding boxes# radians = a * math.pi / 180# reduction = max(abs(math.sin(radians)), abs(math.cos(radians))) ** 0.5# x = (xy[:, 2] + xy[:, 0]) / 2# y = (xy[:, 3] + xy[:, 1]) / 2# w = (xy[:, 2] - xy[:, 0]) * reduction# h = (xy[:, 3] - xy[:, 1]) * reduction# xy = np.concatenate((x - w / 2, y - h / 2, x + w / 2, y + h / 2)).reshape(4, n).T# clip boxes 将超出图像边界的真实框的坐标赋予0或图像边长xy[:, [0, 2]] = xy[:, [0, 2]].clip(0, width)xy[:, [1, 3]] = xy[:, [1, 3]].clip(0, height)# filter candidates  筛选掉过于狭窄，高或宽小于2, 处理之后的真实框的面积要比处理之前真实框的面积<=0.1的真实框i = box_candidates(box1=targets[:, 1:5].T * s, box2=xy.T)targets = targets[i]targets[:, 1:5] = xy[i]return img, targets

letterbox 自适应缩放图片

def letterbox(img, new_shape=(640, 640), color=(114, 114, 114), auto=True, scaleFill=False, scaleup=True):# Resize image to a 32-pixel-multiple rectangle https://github.com/ultralytics/yolov3/issues/232# 调整图片大小，达到32的最小倍数shape = img.shape[:2]  # current shape [height, width]if isinstance(new_shape, int):  # [height, width]new_shape = (new_shape, new_shape)# Scale ratio (new / old)   选择最小的缩放系数r = min(new_shape[0] / shape[0], new_shape[1] / shape[1])"""缩放(resize)到输入大小img_size的时候，如果没有设置上采样的话，则只进行下采样因为上采样图片会让图片模糊，对训练不友好影响性能。"""if not scaleup:  # only scale down, do not scale up (for better test mAP)r = min(r, 1.0)# Compute paddingratio = r, r  # width, height ratiosnew_unpad = int(round(shape[1] * r)), int(round(shape[0] * r))      # [width, height]dw, dh = new_shape[1] - new_unpad[0], new_shape[0] - new_unpad[1]  # wh paddingif auto:  # minimum rectangle 最小矩形填充dw, dh = np.mod(dw, 32), np.mod(dh, 32)  # wh paddingelif scaleFill:  # stretch  直接resize为img_size大小，任由图片拉伸压缩dw, dh = 0.0, 0.0new_unpad = (new_shape[1], new_shape[0])ratio = new_shape[1] / shape[1], new_shape[0] / shape[0]  # width, height ratios# 图像两边需要填充的宽度dw /= 2  # divide padding into 2 sidesdh /= 2if shape[::-1] != new_unpad:  # resizeimg = cv2.resize(img, new_unpad, interpolation=cv2.INTER_LINEAR)top, bottom = int(round(dh - 0.1)), int(round(dh + 0.1))left, right = int(round(dw - 0.1)), int(round(dw + 0.1))# 进行填充img = cv2.copyMakeBorder(img, top, bottom, left, right, cv2.BORDER_CONSTANT, value=color)  # add borderreturn img, ratio, (dw, dh)

YOLOV5源码解读（数据集加载和增强）相关推荐

利用yolov5训练自己的数据集； yolov5的安装与使用； yolov5源码解读
*免责声明: 1\此方法仅提供参考 2\搬了其他博主的操作方法,以贴上路径. 3* 场景一:Anconda环境基本操作场景二:yolov5的使用场景三:yolo v5训练自己的数据集场景四:yo ...
【Android 插件化】Hook 插件化框架 ( 从源码角度分析加载资源流程 | Hook 点选择 | 资源冲突解决方案 )
Android 插件化系列文章目录 [Android 插件化]插件化简介 ( 组件化与插件化 ) [Android 插件化]插件化原理 ( JVM 内存数据 | 类加载流程 ) [Android 插件 ...
Mybatis3源码分析(05)-加载Configuration-加载MappedStatement
2019独角兽企业重金招聘Python工程师标准>>> Mybatis3源码分析(05)-加载Configuration-加载MappedStatement 博客分类: java m ...
springboot集成mybatis源码分析-启动加载mybatis过程（二）
springboot集成mybatis源码分析-启动加载mybatis过程(二) 1.springboot项目最核心的就是自动加载配置,该功能则依赖的是一个注解@SpringBootApplicati ...
Spring源码——bean的加载
前言内容主要参考自<Spring源码深度解析>一书,算是读书笔记或是原书的补充.进入正文后可能会引来各种不适,毕竟阅读源码是件极其痛苦的事情. 本文主要涉及书中第五章的部分,依照书中内容 ...
【转载保存】修改IK分词器源码实现动态加载词典
链接:http://www.gongstring.com/portal/article/index/id/59.html 当前IKAnalyzer从发布最后一个版本后就一直没有再更新,使用过程中,经常 ...
从 RequireJs 源码剖析脚本加载原理
引言俗话说的好,不喜欢研究原理的程序员不是好的程序员,不喜欢读源码的程序员不是好的 jser.这两天看到了有关前端模块化的问题,才发现 JavaScript 社区为了前端工程化真是煞费苦心.今天研究 ...
skywalking源码--agent配置加载
本源码来自于skywalking-agent 8.9.0版本本节主要讲解skywalking-agent的配置文件加载流程,该实现在 apm-sniffer模块的apm-agent-core 模块里 ...
Spring源码分析4---IOC加载过程补充总结
原文出自:http://cmsblogs.com IOC 之获取验证模型 DTD 与 XSD 的区别 DTD(Document Type Definition),即文档类型定义,为 XML 文件的验 ...

YOLOV5源码解读（数据集加载和增强）