sun15684712291/aicheckv2-api
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

完成训练模块的转移

Browse Source
This commit is contained in:
sunyg
2025-04-17 11:03:05 +08:00
parent 4439687870
commit 74e8f0d415
188 changed files with 32931 additions and 70 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
yolov5/utils/segment/__init__.py Normal file
View File

@ -0,0 +1 @@
# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license

92
yolov5/utils/segment/augmentations.py Normal file
View File

@ -0,0 +1,92 @@
# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
"""Image augmentation functions."""
import math
import random
import cv2
import numpy as np
from ..augmentations import box_candidates
from ..general import resample_segments, segment2box
def mixup(im, labels, segments, im2, labels2, segments2):
"""
Applies MixUp augmentation blending two images, labels, and segments with a random ratio.
See https://arxiv.org/pdf/1710.09412.pdf
"""
r = np.random.beta(32.0, 32.0) # mixup ratio, alpha=beta=32.0
im = (im * r + im2 * (1 - r)).astype(np.uint8)
labels = np.concatenate((labels, labels2), 0)
segments = np.concatenate((segments, segments2), 0)
return im, labels, segments
def random_perspective(
im, targets=(), segments=(), degrees=10, translate=0.1, scale=0.1, 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]
"""Applies random perspective, rotation, scale, shear, and translation augmentations to an image and targets."""
height = im.shape[0] + border[0] * 2 # shape(h,w,c)
width = im.shape[1] + border[1] * 2
# Center
C = np.eye(3)
C[0, 2] = -im.shape[1] / 2 # x translation (pixels)
C[1, 2] = -im.shape[0] / 2 # y translation (pixels)
# Perspective
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
R = np.eye(3)
a = random.uniform(-degrees, degrees)
# a += random.choice([-180, -90, 0, 90]) # add 90deg rotations to small rotations
s = random.uniform(1 - scale, 1 + scale)
# s = 2 ** random.uniform(-scale, scale)
R[:2] = cv2.getRotationMatrix2D(angle=a, center=(0, 0), scale=s)
# Shear
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
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
M = T @ S @ R @ P @ C # order of operations (right to left) is IMPORTANT
if (border[0] != 0) or (border[1] != 0) or (M != np.eye(3)).any(): # image changed
if perspective:
im = cv2.warpPerspective(im, M, dsize=(width, height), borderValue=(114, 114, 114))
else: # affine
im = cv2.warpAffine(im, M[:2], dsize=(width, height), borderValue=(114, 114, 114))
new_segments = []
if n := len(targets):
new = np.zeros((n, 4))
segments = resample_segments(segments) # upsample
for i, segment in enumerate(segments):
xy = np.ones((len(segment), 3))
xy[:, :2] = segment
xy = xy @ M.T # transform
xy = xy[:, :2] / xy[:, 2:3] if perspective else xy[:, :2] # perspective rescale or affine
# clip
new[i] = segment2box(xy, width, height)
new_segments.append(xy)
# filter candidates
i = box_candidates(box1=targets[:, 1:5].T * s, box2=new.T, area_thr=0.01)
targets = targets[i]
targets[:, 1:5] = new[i]
new_segments = np.array(new_segments)[i]
return im, targets, new_segments

366
yolov5/utils/segment/dataloaders.py Normal file
View File

@ -0,0 +1,366 @@
# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
"""Dataloaders."""
import os
import random
import cv2
import numpy as np
import torch
from torch.utils.data import DataLoader
from ..augmentations import augment_hsv, copy_paste, letterbox
from ..dataloaders import InfiniteDataLoader, LoadImagesAndLabels, SmartDistributedSampler, seed_worker
from ..general import LOGGER, xyn2xy, xywhn2xyxy, xyxy2xywhn
from ..torch_utils import torch_distributed_zero_first
from .augmentations import mixup, random_perspective
RANK = int(os.getenv("RANK", -1))
def create_dataloader(
path,
imgsz,
batch_size,
stride,
single_cls=False,
hyp=None,
augment=False,
cache=False,
pad=0.0,
rect=False,
rank=-1,
workers=8,
image_weights=False,
quad=False,
prefix="",
shuffle=False,
mask_downsample_ratio=1,
overlap_mask=False,
seed=0,
):
"""Creates a dataloader for training, validating, or testing YOLO models with various dataset options."""
if rect and shuffle:
LOGGER.warning("WARNING ⚠️ --rect is incompatible with DataLoader shuffle, setting shuffle=False")
shuffle = False
with torch_distributed_zero_first(rank): # init dataset *.cache only once if DDP
dataset = LoadImagesAndLabelsAndMasks(
path,
imgsz,
batch_size,
augment=augment, # augmentation
hyp=hyp, # hyperparameters
rect=rect, # rectangular batches
cache_images=cache,
single_cls=single_cls,
stride=int(stride),
pad=pad,
image_weights=image_weights,
prefix=prefix,
downsample_ratio=mask_downsample_ratio,
overlap=overlap_mask,
rank=rank,
)
batch_size = min(batch_size, len(dataset))
nd = torch.cuda.device_count() # number of CUDA devices
nw = min([os.cpu_count() // max(nd, 1), batch_size if batch_size > 1 else 0, workers]) # number of workers
sampler = None if rank == -1 else SmartDistributedSampler(dataset, shuffle=shuffle)
loader = DataLoader if image_weights else InfiniteDataLoader # only DataLoader allows for attribute updates
generator = torch.Generator()
generator.manual_seed(6148914691236517205 + seed + RANK)
return loader(
dataset,
batch_size=batch_size,
shuffle=shuffle and sampler is None,
num_workers=nw,
sampler=sampler,
drop_last=quad,
pin_memory=True,
collate_fn=LoadImagesAndLabelsAndMasks.collate_fn4 if quad else LoadImagesAndLabelsAndMasks.collate_fn,
worker_init_fn=seed_worker,
generator=generator,
), dataset
class LoadImagesAndLabelsAndMasks(LoadImagesAndLabels): # for training/testing
"""Loads images, labels, and segmentation masks for training and testing YOLO models with augmentation support."""
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,
min_items=0,
prefix="",
downsample_ratio=1,
overlap=False,
rank=-1,
seed=0,
):
"""Initializes the dataset with image, label, and mask loading capabilities for training/testing."""
super().__init__(
path,
img_size,
batch_size,
augment,
hyp,
rect,
image_weights,
cache_images,
single_cls,
stride,
pad,
min_items,
prefix,
rank,
seed,
)
self.downsample_ratio = downsample_ratio
self.overlap = overlap
def __getitem__(self, index):
"""Returns a transformed item from the dataset at the specified index, handling indexing and image weighting."""
index = self.indices[index] # linear, shuffled, or image_weights
hyp = self.hyp
if mosaic := self.mosaic and random.random() < hyp["mosaic"]:
# Load mosaic
img, labels, segments = self.load_mosaic(index)
shapes = None
# MixUp augmentation
if random.random() < hyp["mixup"]:
img, labels, segments = mixup(img, labels, segments, *self.load_mosaic(random.randint(0, self.n - 1)))
else:
# Load image
img, (h0, w0), (h, w) = self.load_image(index)
# 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)
shapes = (h0, w0), ((h / h0, w / w0), pad) # for COCO mAP rescaling
labels = self.labels[index].copy()
# [array, array, ....], array.shape=(num_points, 2), xyxyxyxy
segments = self.segments[index].copy()
if len(segments):
for i_s in range(len(segments)):
segments[i_s] = xyn2xy(
segments[i_s],
ratio[0] * w,
ratio[1] * h,
padw=pad[0],
padh=pad[1],
)
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, segments = random_perspective(
img,
labels,
segments=segments,
degrees=hyp["degrees"],
translate=hyp["translate"],
scale=hyp["scale"],
shear=hyp["shear"],
perspective=hyp["perspective"],
)
nl = len(labels) # number of labels
masks = []
if nl:
labels[:, 1:5] = xyxy2xywhn(labels[:, 1:5], w=img.shape[1], h=img.shape[0], clip=True, eps=1e-3)
if self.overlap:
masks, sorted_idx = polygons2masks_overlap(
img.shape[:2], segments, downsample_ratio=self.downsample_ratio
)
masks = masks[None] # (640, 640) -> (1, 640, 640)
labels = labels[sorted_idx]
else:
masks = polygons2masks(img.shape[:2], segments, color=1, downsample_ratio=self.downsample_ratio)
masks = (
torch.from_numpy(masks)
if len(masks)
else torch.zeros(
1 if self.overlap else nl, img.shape[0] // self.downsample_ratio, img.shape[1] // self.downsample_ratio
)
)
# TODO: albumentations support
if self.augment:
# Albumentations
# there are some augmentation that won't change boxes and masks,
# so just be it for now.
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
if random.random() < hyp["flipud"]:
img = np.flipud(img)
if nl:
labels[:, 2] = 1 - labels[:, 2]
masks = torch.flip(masks, dims=[1])
# Flip left-right
if random.random() < hyp["fliplr"]:
img = np.fliplr(img)
if nl:
labels[:, 1] = 1 - labels[:, 1]
masks = torch.flip(masks, dims=[2])
# Cutouts # labels = cutout(img, labels, p=0.5)
labels_out = torch.zeros((nl, 6))
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)
return (torch.from_numpy(img), labels_out, self.im_files[index], shapes, masks)
def load_mosaic(self, index):
"""Loads 1 image + 3 random images into a 4-image YOLOv5 mosaic, adjusting labels and segments accordingly."""
labels4, segments4 = [], []
s = self.img_size
yc, xc = (int(random.uniform(-x, 2 * s + x)) for x in self.mosaic_border) # mosaic center x, y
# 3 additional image indices
indices = [index] + random.choices(self.indices, k=3) # 3 additional image indices
for i, index in enumerate(indices):
# Load image
img, _, (h, w) = self.load_image(index)
# place img in img4
if i == 0: # top left
img4 = np.full((s * 2, s * 2, img.shape[2]), 114, dtype=np.uint8) # base image with 4 tiles
x1a, y1a, x2a, y2a = max(xc - w, 0), max(yc - h, 0), xc, yc # xmin, ymin, xmax, ymax (large image)
x1b, y1b, x2b, y2b = w - (x2a - x1a), h - (y2a - y1a), w, h # xmin, ymin, xmax, ymax (small image)
elif i == 1: # top right
x1a, y1a, x2a, y2a = xc, max(yc - h, 0), min(xc + w, s * 2), yc
x1b, y1b, x2b, y2b = 0, h - (y2a - y1a), min(w, x2a - x1a), h
elif i == 2: # bottom left
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
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)
img4[y1a:y2a, x1a:x2a] = img[y1b:y2b, x1b:x2b] # img4[ymin:ymax, xmin:xmax]
padw = x1a - x1b
padh = y1a - y1b
labels, segments = self.labels[index].copy(), self.segments[index].copy()
if labels.size:
labels[:, 1:] = xywhn2xyxy(labels[:, 1:], w, h, padw, padh) # normalized xywh to pixel xyxy format
segments = [xyn2xy(x, w, h, padw, padh) for x in segments]
labels4.append(labels)
segments4.extend(segments)
# Concat/clip labels
labels4 = np.concatenate(labels4, 0)
for x in (labels4[:, 1:], *segments4):
np.clip(x, 0, 2 * s, out=x) # clip when using random_perspective()
# img4, labels4 = replicate(img4, labels4) # replicate
# Augment
img4, labels4, segments4 = copy_paste(img4, labels4, segments4, p=self.hyp["copy_paste"])
img4, labels4, segments4 = random_perspective(
img4,
labels4,
segments4,
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 remove
return img4, labels4, segments4
@staticmethod
def collate_fn(batch):
"""Custom collation function for DataLoader, batches images, labels, paths, shapes, and segmentation masks."""
img, label, path, shapes, masks = zip(*batch) # transposed
batched_masks = torch.cat(masks, 0)
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, batched_masks
def polygon2mask(img_size, polygons, color=1, downsample_ratio=1):
"""
Args:
img_size (tuple): The image size.
polygons (np.ndarray): [N, M], N is the number of polygons,
M is the number of points(Be divided by 2).
"""
mask = np.zeros(img_size, dtype=np.uint8)
polygons = np.asarray(polygons)
polygons = polygons.astype(np.int32)
shape = polygons.shape
polygons = polygons.reshape(shape[0], -1, 2)
cv2.fillPoly(mask, polygons, color=color)
nh, nw = (img_size[0] // downsample_ratio, img_size[1] // downsample_ratio)
# NOTE: fillPoly firstly then resize is trying the keep the same way
# of loss calculation when mask-ratio=1.
mask = cv2.resize(mask, (nw, nh))
return mask
def polygons2masks(img_size, polygons, color, downsample_ratio=1):
"""
Args:
img_size (tuple): The image size.
polygons (list[np.ndarray]): each polygon is [N, M],
N is the number of polygons,
M is the number of points(Be divided by 2).
"""
masks = []
for si in range(len(polygons)):
mask = polygon2mask(img_size, [polygons[si].reshape(-1)], color, downsample_ratio)
masks.append(mask)
return np.array(masks)
def polygons2masks_overlap(img_size, segments, downsample_ratio=1):
"""Return a (640, 640) overlap mask."""
masks = np.zeros(
(img_size[0] // downsample_ratio, img_size[1] // downsample_ratio),
dtype=np.int32 if len(segments) > 255 else np.uint8,
)
areas = []
ms = []
for si in range(len(segments)):
mask = polygon2mask(
img_size,
[segments[si].reshape(-1)],
downsample_ratio=downsample_ratio,
color=1,
)
ms.append(mask)
areas.append(mask.sum())
areas = np.asarray(areas)
index = np.argsort(-areas)
ms = np.array(ms)[index]
for i in range(len(segments)):
mask = ms[i] * (i + 1)
masks = masks + mask
masks = np.clip(masks, a_min=0, a_max=i + 1)
return masks, index

160
yolov5/utils/segment/general.py Normal file
View File

@ -0,0 +1,160 @@
# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
import cv2
import numpy as np
import torch
import torch.nn.functional as F
def crop_mask(masks, boxes):
"""
"Crop" predicted masks by zeroing out everything not in the predicted bbox. Vectorized by Chong (thanks Chong).
Args:
- masks should be a size [n, h, w] tensor of masks
- boxes should be a size [n, 4] tensor of bbox coords in relative point form
"""
n, h, w = masks.shape
x1, y1, x2, y2 = torch.chunk(boxes[:, :, None], 4, 1) # x1 shape(1,1,n)
r = torch.arange(w, device=masks.device, dtype=x1.dtype)[None, None, :] # rows shape(1,w,1)
c = torch.arange(h, device=masks.device, dtype=x1.dtype)[None, :, None] # cols shape(h,1,1)
return masks * ((r >= x1) * (r < x2) * (c >= y1) * (c < y2))
def process_mask_upsample(protos, masks_in, bboxes, shape):
"""
Crop after upsample.
protos: [mask_dim, mask_h, mask_w]
masks_in: [n, mask_dim], n is number of masks after nms
bboxes: [n, 4], n is number of masks after nms
shape: input_image_size, (h, w).
return: h, w, n
"""
c, mh, mw = protos.shape # CHW
masks = (masks_in @ protos.float().view(c, -1)).sigmoid().view(-1, mh, mw)
masks = F.interpolate(masks[None], shape, mode="bilinear", align_corners=False)[0] # CHW
masks = crop_mask(masks, bboxes) # CHW
return masks.gt_(0.5)
def process_mask(protos, masks_in, bboxes, shape, upsample=False):
"""
Crop before upsample.
proto_out: [mask_dim, mask_h, mask_w]
out_masks: [n, mask_dim], n is number of masks after nms
bboxes: [n, 4], n is number of masks after nms
shape:input_image_size, (h, w).
return: h, w, n
"""
c, mh, mw = protos.shape # CHW
ih, iw = shape
masks = (masks_in @ protos.float().view(c, -1)).sigmoid().view(-1, mh, mw) # CHW
downsampled_bboxes = bboxes.clone()
downsampled_bboxes[:, 0] *= mw / iw
downsampled_bboxes[:, 2] *= mw / iw
downsampled_bboxes[:, 3] *= mh / ih
downsampled_bboxes[:, 1] *= mh / ih
masks = crop_mask(masks, downsampled_bboxes) # CHW
if upsample:
masks = F.interpolate(masks[None], shape, mode="bilinear", align_corners=False)[0] # CHW
return masks.gt_(0.5)
def process_mask_native(protos, masks_in, bboxes, shape):
"""
Crop after upsample.
protos: [mask_dim, mask_h, mask_w]
masks_in: [n, mask_dim], n is number of masks after nms
bboxes: [n, 4], n is number of masks after nms
shape: input_image_size, (h, w).
return: h, w, n
"""
c, mh, mw = protos.shape # CHW
masks = (masks_in @ protos.float().view(c, -1)).sigmoid().view(-1, mh, mw)
gain = min(mh / shape[0], mw / shape[1]) # gain = old / new
pad = (mw - shape[1] * gain) / 2, (mh - shape[0] * gain) / 2 # wh padding
top, left = int(pad[1]), int(pad[0]) # y, x
bottom, right = int(mh - pad[1]), int(mw - pad[0])
masks = masks[:, top:bottom, left:right]
masks = F.interpolate(masks[None], shape, mode="bilinear", align_corners=False)[0] # CHW
masks = crop_mask(masks, bboxes) # CHW
return masks.gt_(0.5)
def scale_image(im1_shape, masks, im0_shape, ratio_pad=None):
"""
img1_shape: model input shape, [h, w]
img0_shape: origin pic shape, [h, w, 3]
masks: [h, w, num].
"""
# Rescale coordinates (xyxy) from im1_shape to im0_shape
if ratio_pad is None: # calculate from im0_shape
gain = min(im1_shape[0] / im0_shape[0], im1_shape[1] / im0_shape[1]) # gain = old / new
pad = (im1_shape[1] - im0_shape[1] * gain) / 2, (im1_shape[0] - im0_shape[0] * gain) / 2 # wh padding
else:
pad = ratio_pad[1]
top, left = int(pad[1]), int(pad[0]) # y, x
bottom, right = int(im1_shape[0] - pad[1]), int(im1_shape[1] - pad[0])
if len(masks.shape) < 2:
raise ValueError(f'"len of masks shape" should be 2 or 3, but got {len(masks.shape)}')
masks = masks[top:bottom, left:right]
# masks = masks.permute(2, 0, 1).contiguous()
# masks = F.interpolate(masks[None], im0_shape[:2], mode='bilinear', align_corners=False)[0]
# masks = masks.permute(1, 2, 0).contiguous()
masks = cv2.resize(masks, (im0_shape[1], im0_shape[0]))
if len(masks.shape) == 2:
masks = masks[:, :, None]
return masks
def mask_iou(mask1, mask2, eps=1e-7):
"""
mask1: [N, n] m1 means number of predicted objects
mask2: [M, n] m2 means number of gt objects
Note: n means image_w x image_h.
return: masks iou, [N, M]
"""
intersection = torch.matmul(mask1, mask2.t()).clamp(0)
union = (mask1.sum(1)[:, None] + mask2.sum(1)[None]) - intersection # (area1 + area2) - intersection
return intersection / (union + eps)
def masks_iou(mask1, mask2, eps=1e-7):
"""
mask1: [N, n] m1 means number of predicted objects
mask2: [N, n] m2 means number of gt objects
Note: n means image_w x image_h.
return: masks iou, (N, )
"""
intersection = (mask1 * mask2).sum(1).clamp(0) # (N, )
union = (mask1.sum(1) + mask2.sum(1))[None] - intersection # (area1 + area2) - intersection
return intersection / (union + eps)
def masks2segments(masks, strategy="largest"):
"""Converts binary (n,160,160) masks to polygon segments with options for concatenation or selecting the largest
segment.
"""
segments = []
for x in masks.int().cpu().numpy().astype("uint8"):
c = cv2.findContours(x, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)[0]
if c:
if strategy == "concat": # concatenate all segments
c = np.concatenate([x.reshape(-1, 2) for x in c])
elif strategy == "largest": # select largest segment
c = np.array(c[np.array([len(x) for x in c]).argmax()]).reshape(-1, 2)
else:
c = np.zeros((0, 2)) # no segments found
segments.append(c.astype("float32"))
return segments

197
yolov5/utils/segment/loss.py Normal file
View File

@ -0,0 +1,197 @@
# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
import torch
import torch.nn as nn
import torch.nn.functional as F
from ..general import xywh2xyxy
from ..loss import FocalLoss, smooth_BCE
from ..metrics import bbox_iou
from ..torch_utils import de_parallel
from .general import crop_mask
class ComputeLoss:
"""Computes the YOLOv5 model's loss components including classification, objectness, box, and mask losses."""
def __init__(self, model, autobalance=False, overlap=False):
"""Initializes the compute loss function for YOLOv5 models with options for autobalancing and overlap
handling.
"""
self.sort_obj_iou = False
self.overlap = overlap
device = next(model.parameters()).device # get model device
h = model.hyp # hyperparameters
# Define criteria
BCEcls = nn.BCEWithLogitsLoss(pos_weight=torch.tensor([h["cls_pw"]], device=device))
BCEobj = nn.BCEWithLogitsLoss(pos_weight=torch.tensor([h["obj_pw"]], device=device))
# Class label smoothing https://arxiv.org/pdf/1902.04103.pdf eqn 3
self.cp, self.cn = smooth_BCE(eps=h.get("label_smoothing", 0.0)) # positive, negative BCE targets
# Focal loss
g = h["fl_gamma"] # focal loss gamma
if g > 0:
BCEcls, BCEobj = FocalLoss(BCEcls, g), FocalLoss(BCEobj, g)
m = de_parallel(model).model[-1] # Detect() module
self.balance = {3: [4.0, 1.0, 0.4]}.get(m.nl, [4.0, 1.0, 0.25, 0.06, 0.02]) # P3-P7
self.ssi = list(m.stride).index(16) if autobalance else 0 # stride 16 index
self.BCEcls, self.BCEobj, self.gr, self.hyp, self.autobalance = BCEcls, BCEobj, 1.0, h, autobalance
self.na = m.na # number of anchors
self.nc = m.nc # number of classes
self.nl = m.nl # number of layers
self.nm = m.nm # number of masks
self.anchors = m.anchors
self.device = device
def __call__(self, preds, targets, masks): # predictions, targets, model
"""Evaluates YOLOv5 model's loss for given predictions, targets, and masks; returns total loss components."""
p, proto = preds
bs, nm, mask_h, mask_w = proto.shape # batch size, number of masks, mask height, mask width
lcls = torch.zeros(1, device=self.device)
lbox = torch.zeros(1, device=self.device)
lobj = torch.zeros(1, device=self.device)
lseg = torch.zeros(1, device=self.device)
tcls, tbox, indices, anchors, tidxs, xywhn = self.build_targets(p, targets) # targets
# Losses
for i, pi in enumerate(p): # layer index, layer predictions
b, a, gj, gi = indices[i] # image, anchor, gridy, gridx
tobj = torch.zeros(pi.shape[:4], dtype=pi.dtype, device=self.device) # target obj
if n := b.shape[0]:
pxy, pwh, _, pcls, pmask = pi[b, a, gj, gi].split((2, 2, 1, self.nc, nm), 1) # subset of predictions
# Box regression
pxy = pxy.sigmoid() * 2 - 0.5
pwh = (pwh.sigmoid() * 2) ** 2 * anchors[i]
pbox = torch.cat((pxy, pwh), 1) # predicted box
iou = bbox_iou(pbox, tbox[i], CIoU=True).squeeze() # iou(prediction, target)
lbox += (1.0 - iou).mean() # iou loss
# Objectness
iou = iou.detach().clamp(0).type(tobj.dtype)
if self.sort_obj_iou:
j = iou.argsort()
b, a, gj, gi, iou = b[j], a[j], gj[j], gi[j], iou[j]
if self.gr < 1:
iou = (1.0 - self.gr) + self.gr * iou
tobj[b, a, gj, gi] = iou # iou ratio
# Classification
if self.nc > 1: # cls loss (only if multiple classes)
t = torch.full_like(pcls, self.cn, device=self.device) # targets
t[range(n), tcls[i]] = self.cp
lcls += self.BCEcls(pcls, t) # BCE
# Mask regression
if tuple(masks.shape[-2:]) != (mask_h, mask_w): # downsample
masks = F.interpolate(masks[None], (mask_h, mask_w), mode="nearest")[0]
marea = xywhn[i][:, 2:].prod(1) # mask width, height normalized
mxyxy = xywh2xyxy(xywhn[i] * torch.tensor([mask_w, mask_h, mask_w, mask_h], device=self.device))
for bi in b.unique():
j = b == bi # matching index
if self.overlap:
mask_gti = torch.where(masks[bi][None] == tidxs[i][j].view(-1, 1, 1), 1.0, 0.0)
else:
mask_gti = masks[tidxs[i]][j]
lseg += self.single_mask_loss(mask_gti, pmask[j], proto[bi], mxyxy[j], marea[j])
obji = self.BCEobj(pi[..., 4], tobj)
lobj += obji * self.balance[i] # obj loss
if self.autobalance:
self.balance[i] = self.balance[i] * 0.9999 + 0.0001 / obji.detach().item()
if self.autobalance:
self.balance = [x / self.balance[self.ssi] for x in self.balance]
lbox *= self.hyp["box"]
lobj *= self.hyp["obj"]
lcls *= self.hyp["cls"]
lseg *= self.hyp["box"] / bs
loss = lbox + lobj + lcls + lseg
return loss * bs, torch.cat((lbox, lseg, lobj, lcls)).detach()
def single_mask_loss(self, gt_mask, pred, proto, xyxy, area):
"""Calculates and normalizes single mask loss for YOLOv5 between predicted and ground truth masks."""
pred_mask = (pred @ proto.view(self.nm, -1)).view(-1, *proto.shape[1:]) # (n,32) @ (32,80,80) -> (n,80,80)
loss = F.binary_cross_entropy_with_logits(pred_mask, gt_mask, reduction="none")
return (crop_mask(loss, xyxy).mean(dim=(1, 2)) / area).mean()
def build_targets(self, p, targets):
"""Prepares YOLOv5 targets for loss computation; inputs targets (image, class, x, y, w, h), output target
classes/boxes.
"""
na, nt = self.na, targets.shape[0] # number of anchors, targets
tcls, tbox, indices, anch, tidxs, xywhn = [], [], [], [], [], []
gain = torch.ones(8, device=self.device) # normalized to gridspace gain
ai = torch.arange(na, device=self.device).float().view(na, 1).repeat(1, nt) # same as .repeat_interleave(nt)
if self.overlap:
batch = p[0].shape[0]
ti = []
for i in range(batch):
num = (targets[:, 0] == i).sum() # find number of targets of each image
ti.append(torch.arange(num, device=self.device).float().view(1, num).repeat(na, 1) + 1) # (na, num)
ti = torch.cat(ti, 1) # (na, nt)
else:
ti = torch.arange(nt, device=self.device).float().view(1, nt).repeat(na, 1)
targets = torch.cat((targets.repeat(na, 1, 1), ai[..., None], ti[..., None]), 2) # append anchor indices
g = 0.5 # bias
off = (
torch.tensor(
[
[0, 0],
[1, 0],
[0, 1],
[-1, 0],
[0, -1], # j,k,l,m
# [1, 1], [1, -1], [-1, 1], [-1, -1], # jk,jm,lk,lm
],
device=self.device,
).float()
* g
) # offsets
for i in range(self.nl):
anchors, shape = self.anchors[i], p[i].shape
gain[2:6] = torch.tensor(shape)[[3, 2, 3, 2]] # xyxy gain
# Match targets to anchors
t = targets * gain # shape(3,n,7)
if nt:
# Matches
r = t[..., 4:6] / anchors[:, None] # wh ratio
j = torch.max(r, 1 / r).max(2)[0] < self.hyp["anchor_t"] # compare
# j = wh_iou(anchors, t[:, 4:6]) > model.hyp['iou_t'] # iou(3,n)=wh_iou(anchors(3,2), gwh(n,2))
t = t[j] # filter
# Offsets
gxy = t[:, 2:4] # grid xy
gxi = gain[[2, 3]] - gxy # inverse
j, k = ((gxy % 1 < g) & (gxy > 1)).T
l, m = ((gxi % 1 < g) & (gxi > 1)).T
j = torch.stack((torch.ones_like(j), j, k, l, m))
t = t.repeat((5, 1, 1))[j]
offsets = (torch.zeros_like(gxy)[None] + off[:, None])[j]
else:
t = targets[0]
offsets = 0
# Define
bc, gxy, gwh, at = t.chunk(4, 1) # (image, class), grid xy, grid wh, anchors
(a, tidx), (b, c) = at.long().T, bc.long().T # anchors, image, class
gij = (gxy - offsets).long()
gi, gj = gij.T # grid indices
# Append
indices.append((b, a, gj.clamp_(0, shape[2] - 1), gi.clamp_(0, shape[3] - 1))) # image, anchor, grid
tbox.append(torch.cat((gxy - gij, gwh), 1)) # box
anch.append(anchors[a]) # anchors
tcls.append(c) # class
tidxs.append(tidx)
xywhn.append(torch.cat((gxy, gwh), 1) / gain[2:6]) # xywh normalized
return tcls, tbox, indices, anch, tidxs, xywhn

225
yolov5/utils/segment/metrics.py Normal file
View File

@ -0,0 +1,225 @@
# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
"""Model validation metrics."""
import numpy as np
from ..metrics import ap_per_class
def fitness(x):
"""Evaluates model fitness by a weighted sum of 8 metrics, `x`: [N,8] array, weights: [0.1, 0.9] for mAP and F1."""
w = [0.0, 0.0, 0.1, 0.9, 0.0, 0.0, 0.1, 0.9]
return (x[:, :8] * w).sum(1)
def ap_per_class_box_and_mask(
tp_m,
tp_b,
conf,
pred_cls,
target_cls,
plot=False,
save_dir=".",
names=(),
):
"""
Args:
tp_b: tp of boxes.
tp_m: tp of masks.
other arguments see `func: ap_per_class`.
"""
results_boxes = ap_per_class(
tp_b, conf, pred_cls, target_cls, plot=plot, save_dir=save_dir, names=names, prefix="Box"
)[2:]
results_masks = ap_per_class(
tp_m, conf, pred_cls, target_cls, plot=plot, save_dir=save_dir, names=names, prefix="Mask"
)[2:]
return {
"boxes": {
"p": results_boxes[0],
"r": results_boxes[1],
"ap": results_boxes[3],
"f1": results_boxes[2],
"ap_class": results_boxes[4],
},
"masks": {
"p": results_masks[0],
"r": results_masks[1],
"ap": results_masks[3],
"f1": results_masks[2],
"ap_class": results_masks[4],
},
}
class Metric:
"""Computes performance metrics like precision, recall, F1 score, and average precision for model evaluation."""
def __init__(self) -> None:
"""Initializes performance metric attributes for precision, recall, F1 score, average precision, and class
indices.
"""
self.p = [] # (nc, )
self.r = [] # (nc, )
self.f1 = [] # (nc, )
self.all_ap = [] # (nc, 10)
self.ap_class_index = [] # (nc, )
@property
def ap50(self):
"""
AP@0.5 of all classes.
Return:
(nc, ) or [].
"""
return self.all_ap[:, 0] if len(self.all_ap) else []
@property
def ap(self):
"""AP@0.5:0.95
Return:
(nc, ) or [].
"""
return self.all_ap.mean(1) if len(self.all_ap) else []
@property
def mp(self):
"""
Mean precision of all classes.
Return:
float.
"""
return self.p.mean() if len(self.p) else 0.0
@property
def mr(self):
"""
Mean recall of all classes.
Return:
float.
"""
return self.r.mean() if len(self.r) else 0.0
@property
def map50(self):
"""
Mean AP@0.5 of all classes.
Return:
float.
"""
return self.all_ap[:, 0].mean() if len(self.all_ap) else 0.0
@property
def map(self):
"""
Mean AP@0.5:0.95 of all classes.
Return:
float.
"""
return self.all_ap.mean() if len(self.all_ap) else 0.0
def mean_results(self):
"""Mean of results, return mp, mr, map50, map."""
return (self.mp, self.mr, self.map50, self.map)
def class_result(self, i):
"""Class-aware result, return p[i], r[i], ap50[i], ap[i]."""
return (self.p[i], self.r[i], self.ap50[i], self.ap[i])
def get_maps(self, nc):
"""Calculates and returns mean Average Precision (mAP) for each class given number of classes `nc`."""
maps = np.zeros(nc) + self.map
for i, c in enumerate(self.ap_class_index):
maps[c] = self.ap[i]
return maps
def update(self, results):
"""
Args:
results: tuple(p, r, ap, f1, ap_class).
"""
p, r, all_ap, f1, ap_class_index = results
self.p = p
self.r = r
self.all_ap = all_ap
self.f1 = f1
self.ap_class_index = ap_class_index
class Metrics:
"""Metric for boxes and masks."""
def __init__(self) -> None:
"""Initializes Metric objects for bounding boxes and masks to compute performance metrics in the Metrics
class.
"""
self.metric_box = Metric()
self.metric_mask = Metric()
def update(self, results):
"""
Args:
results: Dict{'boxes': Dict{}, 'masks': Dict{}}.
"""
self.metric_box.update(list(results["boxes"].values()))
self.metric_mask.update(list(results["masks"].values()))
def mean_results(self):
"""Computes and returns the mean results for both box and mask metrics by summing their individual means."""
return self.metric_box.mean_results() + self.metric_mask.mean_results()
def class_result(self, i):
"""Returns the sum of box and mask metric results for a specified class index `i`."""
return self.metric_box.class_result(i) + self.metric_mask.class_result(i)
def get_maps(self, nc):
"""Calculates and returns the sum of mean average precisions (mAPs) for both box and mask metrics for `nc`
classes.
"""
return self.metric_box.get_maps(nc) + self.metric_mask.get_maps(nc)
@property
def ap_class_index(self):
"""Returns the class index for average precision, shared by both box and mask metrics."""
return self.metric_box.ap_class_index
KEYS = [
"train/box_loss",
"train/seg_loss", # train loss
"train/obj_loss",
"train/cls_loss",
"metrics/precision(B)",
"metrics/recall(B)",
"metrics/mAP_0.5(B)",
"metrics/mAP_0.5:0.95(B)", # metrics
"metrics/precision(M)",
"metrics/recall(M)",
"metrics/mAP_0.5(M)",
"metrics/mAP_0.5:0.95(M)", # metrics
"val/box_loss",
"val/seg_loss", # val loss
"val/obj_loss",
"val/cls_loss",
"x/lr0",
"x/lr1",
"x/lr2",
]
BEST_KEYS = [
"best/epoch",
"best/precision(B)",
"best/recall(B)",
"best/mAP_0.5(B)",
"best/mAP_0.5:0.95(B)",
"best/precision(M)",
"best/recall(M)",
"best/mAP_0.5(M)",
"best/mAP_0.5:0.95(M)",
]

152
yolov5/utils/segment/plots.py Normal file
View File

@ -0,0 +1,152 @@
# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
import contextlib
import math
from pathlib import Path
import cv2
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import torch
from .. import threaded
from ..general import xywh2xyxy
from ..plots import Annotator, colors
@threaded
def plot_images_and_masks(images, targets, masks, paths=None, fname="images.jpg", names=None):
"""Plots a grid of images, their labels, and masks with optional resizing and annotations, saving to fname."""
if isinstance(images, torch.Tensor):
images = images.cpu().float().numpy()
if isinstance(targets, torch.Tensor):
targets = targets.cpu().numpy()
if isinstance(masks, torch.Tensor):
masks = masks.cpu().numpy().astype(int)
max_size = 1920 # max image size
max_subplots = 16 # max image subplots, i.e. 4x4
bs, _, h, w = images.shape # batch size, _, height, width
bs = min(bs, max_subplots) # limit plot images
ns = np.ceil(bs**0.5) # number of subplots (square)
if np.max(images[0]) <= 1:
images *= 255 # de-normalise (optional)
# Build Image
mosaic = np.full((int(ns * h), int(ns * w), 3), 255, dtype=np.uint8) # init
for i, im in enumerate(images):
if i == max_subplots: # if last batch has fewer images than we expect
break
x, y = int(w * (i // ns)), int(h * (i % ns)) # block origin
im = im.transpose(1, 2, 0)
mosaic[y : y + h, x : x + w, :] = im
# Resize (optional)
scale = max_size / ns / max(h, w)
if scale < 1:
h = math.ceil(scale * h)
w = math.ceil(scale * w)
mosaic = cv2.resize(mosaic, tuple(int(x * ns) for x in (w, h)))
# Annotate
fs = int((h + w) * ns * 0.01) # font size
annotator = Annotator(mosaic, line_width=round(fs / 10), font_size=fs, pil=True, example=names)
for i in range(i + 1):
x, y = int(w * (i // ns)), int(h * (i % ns)) # block origin
annotator.rectangle([x, y, x + w, y + h], None, (255, 255, 255), width=2) # borders
if paths:
annotator.text([x + 5, y + 5], text=Path(paths[i]).name[:40], txt_color=(220, 220, 220)) # filenames
if len(targets) > 0:
idx = targets[:, 0] == i
ti = targets[idx] # image targets
boxes = xywh2xyxy(ti[:, 2:6]).T
classes = ti[:, 1].astype("int")
labels = ti.shape[1] == 6 # labels if no conf column
conf = None if labels else ti[:, 6] # check for confidence presence (label vs pred)
if boxes.shape[1]:
if boxes.max() <= 1.01: # if normalized with tolerance 0.01
boxes[[0, 2]] *= w # scale to pixels
boxes[[1, 3]] *= h
elif scale < 1: # absolute coords need scale if image scales
boxes *= scale
boxes[[0, 2]] += x
boxes[[1, 3]] += y
for j, box in enumerate(boxes.T.tolist()):
cls = classes[j]
color = colors(cls)
cls = names[cls] if names else cls
if labels or conf[j] > 0.25: # 0.25 conf thresh
label = f"{cls}" if labels else f"{cls} {conf[j]:.1f}"
annotator.box_label(box, label, color=color)
# Plot masks
if len(masks):
if masks.max() > 1.0: # mean that masks are overlap
image_masks = masks[[i]] # (1, 640, 640)
nl = len(ti)
index = np.arange(nl).reshape(nl, 1, 1) + 1
image_masks = np.repeat(image_masks, nl, axis=0)
image_masks = np.where(image_masks == index, 1.0, 0.0)
else:
image_masks = masks[idx]
im = np.asarray(annotator.im).copy()
for j, box in enumerate(boxes.T.tolist()):
if labels or conf[j] > 0.25: # 0.25 conf thresh
color = colors(classes[j])
mh, mw = image_masks[j].shape
if mh != h or mw != w:
mask = image_masks[j].astype(np.uint8)
mask = cv2.resize(mask, (w, h))
mask = mask.astype(bool)
else:
mask = image_masks[j].astype(bool)
with contextlib.suppress(Exception):
im[y : y + h, x : x + w, :][mask] = (
im[y : y + h, x : x + w, :][mask] * 0.4 + np.array(color) * 0.6
)
annotator.fromarray(im)
annotator.im.save(fname) # save
def plot_results_with_masks(file="path/to/results.csv", dir="", best=True):
"""
Plots training results from CSV files, plotting best or last result highlights based on `best` parameter.
Example: from utils.plots import *; plot_results('path/to/results.csv')
"""
save_dir = Path(file).parent if file else Path(dir)
fig, ax = plt.subplots(2, 8, figsize=(18, 6), tight_layout=True)
ax = ax.ravel()
files = list(save_dir.glob("results*.csv"))
assert len(files), f"No results.csv files found in {save_dir.resolve()}, nothing to plot."
for f in files:
try:
data = pd.read_csv(f)
index = np.argmax(
0.9 * data.values[:, 8] + 0.1 * data.values[:, 7] + 0.9 * data.values[:, 12] + 0.1 * data.values[:, 11]
)
s = [x.strip() for x in data.columns]
x = data.values[:, 0]
for i, j in enumerate([1, 2, 3, 4, 5, 6, 9, 10, 13, 14, 15, 16, 7, 8, 11, 12]):
y = data.values[:, j]
# y[y == 0] = np.nan # don't show zero values
ax[i].plot(x, y, marker=".", label=f.stem, linewidth=2, markersize=2)
if best:
# best
ax[i].scatter(index, y[index], color="r", label=f"best:{index}", marker="*", linewidth=3)
ax[i].set_title(s[j] + f"\n{round(y[index], 5)}")
else:
# last
ax[i].scatter(x[-1], y[-1], color="r", label="last", marker="*", linewidth=3)
ax[i].set_title(s[j] + f"\n{round(y[-1], 5)}")
# if j in [8, 9, 10]: # share train and val loss y axes
# ax[i].get_shared_y_axes().join(ax[i], ax[i - 5])
except Exception as e:
print(f"Warning: Plotting error for {f}: {e}")
ax[1].legend()
fig.savefig(save_dir / "results.png", dpi=200)
plt.close()