yolov5中数据读取并转换成训练格式
主要涉及到四点：
- 数据读取
- cache缓存
- 数据增强与label对应
- 其他一些辅助函数
以下是自己的一些理解，如有纰漏，欢迎交流
class LoadImagesAndLabels(Dataset)
class LoadImagesAndLabels(Dataset):
# YOLOv5 train_loader/val_loader, loads images and labels for training and validation
cache_version = 0.5 # dataset labels *.cache version
# 初始化
def __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, prefix=''):
self.img_size = img_size # 图片大小
self.augment = augment # 是否图片增强
self.hyp = hyp # 超参
self.image_weights = image_weights # 图片权重
self.rect = False if image_weights else rect # 图片长宽比不resize成1
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] # 如果mosaic, 边界
self.stride = stride # 步长
self.path = path # 路径
self.albumentations = Albumentations() if augment else None # 是否使用 Albumentations 库做数据增强
try:
f = [] # image files
for p in path if isinstance(path, list) else [path]:
p = Path(p) # 字符串的路径转成poxis路径 os-agnostic
if p.is_dir(): # dir 匹配所有符合条件的文件，并以list 返回; recursive 是是否采用递归的方式
f += glob.glob(str(p / '**' / '*.*'), recursive=True)
# f = list(p.rglob('**/*.*')) # pathlib
elif p.is_file(): # file(以文件的方式保存路径名，如coco.yaml )
with open(p, 'r') as t:
t = t.read().strip().splitlines() # 以list方式保存每一行路径字符串
parent = str(p.parent) + os.sep
f += [x.replace('./', parent) if x.startswith('./') else x for x in t] # local to global path
# f += [p.parent / x.lstrip(os.sep) for x in t] # local to global path (pathlib)
else:
raise Exception(f'{prefix}{p} does not exist')
self.img_files = sorted([x.replace('/', os.sep) for x in f if x.split('.')[-1].lower() in IMG_FORMATS])
# self.img_files = sorted([x for x in f if x.suffix[1:].lower() in img_formats]) # pathlib
assert self.img_files, f'{prefix}No images found'
except Exception as e:
raise Exception(f'{prefix}Error loading data from {path}: {e}\nSee {HELP_URL}')
# Check cache
self.label_files = img2label_paths(self.img_files) # 将img图片路径转换成对应label路径
cache_path = (p if p.is_file() else Path(self.label_files[0]).parent).with_suffix('.cache')
try:
cache, exists = np.load(cache_path, allow_pickle=True).item(), True # load dict
assert cache['version'] == self.cache_version # same version
assert cache['hash'] == get_hash(self.label_files + self.img_files) # same hash
except: # 否则重新缓存labels
cache, exists = self.cache_labels(cache_path, prefix), False # cache
# Display cache
nf, nm, ne, nc, n = cache.pop('results') # found, missing, empty, corrupted, total
if exists:
d = f"Scanning '{cache_path}' images and labels... {nf} found, {nm} missing, {ne} empty, {nc} corrupted"
tqdm(None, desc=prefix + d, total=n, initial=n) # display cache results
if cache['msgs']:
logging.info('\n'.join(cache['msgs'])) # display warnings
assert nf > 0 or not augment, f'{prefix}No labels in {cache_path}. Can not train without labels. See {HELP_URL}'
# Read cache
[cache.pop(k) for k in ('hash', 'version', 'msgs')] # remove items 将这三个值去掉 留下label,shape,segments
labels, shapes, self.segments = zip(*cache.values()) #返回元组组成的list
self.labels = list(labels)
self.shapes = np.array(shapes, dtype=np.float64)
self.img_files = list(cache.keys()) # update 返回key 组成的list
self.label_files = img2label_paths(cache.keys()) # update 将imgs 路径转成对应的labels 路径
if single_cls: # 如果多类别合并成一个类别， 标签成 0
for x in self.labels:
x[:, 0] = 0
n = len(shapes) # number of images
bi = np.floor(np.arange(n) / batch_size).astype(np.int) # 每张图片属于哪个batch的索引 batch index
nb = bi[-1] + 1 # number of batches
self.batch = bi # batch index of image
self.n = n
self.indices = range(n)
# Rectangular Training
if self.rect:
# 数据样本长宽比不为1
# Sort by aspect ratio
s = self.shapes # wh
ar = s[:, 1] / s[:, 0] # aspect ratio
irect = ar.argsort() # 将长宽比从小到大排序，返回对应的索引
self.img_files = [self.img_files[i] for i in irect] # 将图片按照长宽比从小到大重新排列img_file
self.label_files = [self.label_files[i] for i in irect]# 将图片按照长宽比从小到大重新排列label_file
self.labels = [self.labels[i] for i in irect] # 将图片按照长宽比从小到大重新排列label
self.shapes = s[irect] # wh
ar = ar[irect] # h/w
# Set training image shapes
shapes = [[1, 1]] * nb
for i in range(nb): # 对于每一个batch
ari = ar[bi == i] # 属于该batch的长宽比
mini, maxi = ari.min(), ari.max()
if maxi < 1: # 长宽比[1, <1的值]
shapes[i] = [maxi, 1]
elif mini > 1:
shapes[i] = [1, 1 / mini]
# 对于每个batch，bacth_shape 的长宽比取最大的，或者宽长比最大的那个为整个batch的，同时为了保证上下采样像素点为整数
self.batch_shapes = np.ceil(np.array(shapes) * img_size / stride + pad).astype(np.int) * stride
# Cache images into memory for faster training (WARNING: large datasets may exceed system RAM)
self.imgs, self.img_npy = [None] * n, [None] * n
if cache_images: # 缓存图片
if cache_images == 'disk': # 将图片缓存进disk中
self.im_cache_dir = Path(Path(self.img_files[0]).parent.as_posix() + '_npy') # 图片缓存文件夹
self.img_npy = [self.im_cache_dir / Path(f).with_suffix('.npy').name for f in self.img_files] # 将图片缓存成.npy文件
self.im_cache_dir.mkdir(parents=True, exist_ok=True) # 创建文件夹缓存文件
gb = 0 # Gigabytes of cached images
self.img_hw0, self.img_hw = [None] * n, [None] * n
results = ThreadPool(NUM_THREADS).imap(lambda x: load_image(*x), zip(repeat(self), range(n))) # 多进程加载图片。并将图片resize下，返回resize后的图片和原始长宽比。
pbar = tqdm(enumerate(results), total=n) # 以进度条的型式显示出来
for i, x in pbar:
if cache_images == 'disk':
if not self.img_npy[i].exists():# npy图片文件不存在，重新保存
np.save(self.img_npy[i].as_posix(), x[0])
gb += self.img_npy[i].stat().st_size # 文件大小
else:
self.imgs[i], self.img_hw0[i], self.img_hw[i] = x # im, hw_orig, hw_resized = load_image(self, i)
gb += self.imgs[i].nbytes
pbar.desc = f'{prefix}Caching images ({gb / 1E9:.1f}GB {cache_images})'
pbar.close()
def cache_labels(self, path=Path('./labels.cache'), prefix=''):
# Cache dataset labels, check images and read shapes
x = {} # dict
nm, nf, ne, nc, msgs = 0, 0, 0, 0, [] # number missing, found, empty, corrupt, messages
desc = f"{prefix}Scanning '{path.parent / path.stem}' images and labels..."
with Pool(NUM_THREADS) as pool:
# Pool python 多进程的一个子模块， 可以提供指定数量的进程给用户使用，一般用于需要执行的目标很多，而手动限制进程数量又繁琐时，如果目标少且不用控制进程数量的时候，用Process 类。
pbar = tqdm(pool.imap(verify_image_label, zip(self.img_files, self.label_files, repeat(prefix))),
desc=desc, total=len(self.img_files))
# tqdm 进度条显示；
# pool.imap 输入函数，迭代器，返回iterable
# verify_image_label 验证图片和label 可读，并将label转换成统一格式，拱后面使用。
for im_file, l, shape, segments, nm_f, nf_f, ne_f, nc_f, msg in pbar:
nm += nm_f
nf += nf_f
ne += ne_f
nc += nc_f
if im_file: # 如果图片和label都有， 将标签，图片形状，分割点以字典的方式保存下来
x[im_file] = [l, shape, segments]
if msg: # 如果有miss 或者empty 将对于msg 保存下来
msgs.append(msg)
pbar.desc = f"{desc}{nf} found, {nm} missing, {ne} empty, {nc} corrupted"
pbar.close()
if msgs:
logging.info('\n'.join(msgs))
if nf == 0:
logging.info(f'{prefix}WARNING: No labels found in {path}. See {HELP_URL}')
x['hash'] = get_hash(self.label_files + self.img_files) # 将标签路径和图片路径以hash加密算法保存下来
x['results'] = nf, nm, ne, nc, len(self.img_files) # # found , missing, empty, corrupt
x['msgs'] = msgs # warnings
x['version'] = self.cache_version # cache version
try:
np.save(path, x) # save cache for next time
path.with_suffix('.cache.npy').rename(path) # remove .npy suffix 将numpy保存的label.cache.npy 重命名为label.cache
logging.info(f'{prefix}New cache created: {path}')
except Exception as e:
logging.info(f'{prefix}WARNING: Cache directory {path.parent} is not writeable: {e}') # path not writeable
return x
def __len__(self):
return len(self.img_files) # 返回图片数量
# 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 self
def __getitem__(self, index):
index = self.indices[index] # linear, shuffled, or image_weights
hyp = self.hyp # 超参
mosaic = self.mosaic and random.random() < hyp['mosaic'] # hyp['mosaic'] 取权重，大于随机值mosaic
if mosaic:
# Load mosaic
img, labels = load_mosaic(self, index) # 输出img和标签
shapes = None
# MixUp augmentation mixup在mosaic里面运行
if random.random() < hyp['mixup']:
img, labels = mixup(img, labels, *load_mosaic(self, random.randint(0, self.n - 1))) #*load_mosaic(self, random.randint(0, self.n - 1)) 随机取图片 与之前的img 融合
else:
# Load image
img, (h0, w0), (h, w) = load_image(self, index) # 返回解析的图片、以前的长宽比、resize后的长宽比
# Letterbox
shape = self.batch_shapes[self.batch[index]] if self.rect else self.img_size # final letterboxed shape
img, ratio, pad = letterbox(img, shape, auto=False, scaleup=self.augment) # 将img 以letterbox方式 resize到指定长宽
shapes = (h0, w0), ((h / h0, w / w0), pad) # for COCO mAP rescaling
labels = self.labels[index].copy()
if labels.size: # normalized xywh to pixel xyxy format
labels[:, 1:] = xywhn2xyxy(labels[:, 1:], ratio[0] * w, ratio[1] * h, padw=pad[0], padh=pad[1])
if self.augment:
img, labels = random_perspective(img, labels,
degrees=hyp['degrees'],
translate=hyp['translate'],
scale=hyp['scale'],
shear=hyp['shear'],
perspective=hyp['perspective'])
nl = len(labels) # number of labels
if nl: # 再转成yolo格式的label
labels[:, 1:5] = xyxy2xywhn(labels[:, 1:5], w=img.shape[1], h=img.shape[0], clip=True, eps=1E-3)
if self.augment: # 数据增强
# Albumentations
img, labels = self.albumentations(img, labels)
nl = len(labels) # update after albumentations
# HSV color-space
augment_hsv(img, hgain=hyp['hsv_h'], sgain=hyp['hsv_s'], vgain=hyp['hsv_v'])
# Flip up-down 上下翻转
if random.random() < hyp['flipud']:
img = np.flipud(img)
if nl:
labels[:, 2] = 1 - labels[:, 2]
# Flip left-right 左右翻转
if random.random() < hyp['fliplr']:
img = np.fliplr(img)
if nl:
labels[:, 1] = 1 - labels[:, 1]
# Cutouts
# labels = cutout(img, labels, p=0.5)
labels_out = torch.zeros((nl, 6)) # 6 1-类别标签 4- box 1-对应batch
if nl:
labels_out[:, 1:] = torch.from_numpy(labels)
# Convert
img = img.transpose((2, 0, 1))[::-1] # HWC to CHW, BGR to RGB
img = np.ascontiguousarray(img) # np.ascontiguousarray 将内存不连续的数组，转换成内存连续的数组
return torch.from_numpy(img), labels_out, self.img_files[index], shapes
@staticmethod
def collate_fn(batch):
img, label, path, shapes = zip(*batch) # transposed
for 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
@staticmethod
def collate_fn4(batch):
img, label, path, shapes = zip(*batch) # transposed
n = len(shapes) // 4
img4, label4, path4, shapes4 = [], [], path[:n], shapes[:n]
ho = torch.tensor([[0., 0, 0, 1, 0, 0]])
wo = torch.tensor([[0., 0, 1, 0, 0, 0]])
s = torch.tensor([[1, 1, .5, .5, .5, .5]]) # scale
for i in range(n): # zidane torch.zeros(16,3,720,1280) # BCHW
i *= 4
if random.random() < 0.5:
im = F.interpolate(img[i].unsqueeze(0).float(), scale_factor=2., mode='bilinear', align_corners=False)[
0].type(img[i].type())
l = label[i]
else:
im = torch.cat((torch.cat((img[i], img[i + 1]), 1), torch.cat((img[i + 2], img[i + 3]), 1)), 2)
l = torch.cat((label[i], label[i + 1] + ho, label[i + 2] + wo, label[i + 3] + ho + wo), 0) * s
img4.append(im)
label4.append(l)
for i, l in enumerate(label4):
l[:, 0] = i # add target image index for build_targets()
return torch.stack(img4, 0), torch.cat(label4, 0), path4, shapes4
整体如下：
# YOLOv5