目錄
1、IoU
1.1 什么是IOU
?1.2 IOU代碼
2、GIOU
2.1 為什么提出GIOU
2.2 GIoU代碼
3 DIoU?
3.1 為什么提出DIOU
3.2 DIOU代碼
4 CIOU
4.1 為什么提出CIOU
4.2 CIOU代碼
5 EIOU
5.1 為什么提出EIOU?
5.2 EIOU代碼
6?Wise-IoU
7 YOLOv5中添加GIoU、DIoU、CIoU、EIoU、Wise-IoU損失函數(shù)
1、IoU
1.1 什么是IOU
論文鏈接為：UnitBox: An Advanced Object Detection Network
IoU 的全稱為交并比（Intersection over Union），通過這個(gè)名稱我們大概可以猜到 IoU 的計(jì)算方法。IoU 計(jì)算的是 “預(yù)測的邊框” 和 “真實(shí)的邊框” 的交集和并集的比值。計(jì)算過程如下：
其中，綠色面積代表預(yù)測框B與真實(shí)框A的交集；則
顯而易見，IOU的值越高也說明預(yù)測框與真實(shí)框重合程度越高，代表模型預(yù)測越準(zhǔn)確，反之，IOU越低模型性能越差。?
但是，IOU作為損失函數(shù)會(huì)出現(xiàn)以下問題：
- 如果兩個(gè)框沒有相交，根據(jù)定義，IoU=0，不能度量IoU為零距離遠(yuǎn)近的程度。同時(shí)因?yàn)閘oss=0，沒有梯度回傳，無法進(jìn)行學(xué)習(xí)訓(xùn)練。
- IoU無法精確的反映兩者的重合度大小。如下圖所示，三種情況IoU都相等，但看得出來他們的重合度是不一樣的，左邊的圖回歸的效果最好，右邊的最差。
?1.2 IOU代碼
import numpy as np
def Iou(box1, box2, wh=False):
if wh == False:
xmin1, ymin1, xmax1, ymax1 = box1
xmin2, ymin2, xmax2, ymax2 = box2
else:
xmin1, ymin1 = int(box1[0]-box1[2]/2.0), int(box1[1]-box1[3]/2.0)
xmax1, ymax1 = int(box1[0]+box1[2]/2.0), int(box1[1]+box1[3]/2.0)
xmin2, ymin2 = int(box2[0]-box2[2]/2.0), int(box2[1]-box2[3]/2.0)
xmax2, ymax2 = int(box2[0]+box2[2]/2.0), int(box2[1]+box2[3]/2.0)
# 獲取矩形框交集對應(yīng)的左上角和右下角的坐標(biāo)（intersection）
xx1 = np.max([xmin1, xmin2])
yy1 = np.max([ymin1, ymin2])
xx2 = np.min([xmax1, xmax2])
yy2 = np.min([ymax1, ymax2])
# 計(jì)算兩個(gè)矩形框面積
area1 = (xmax1-xmin1) * (ymax1-ymin1)
area2 = (xmax2-xmin2) * (ymax2-ymin2)
inter_area = (np.max([0, xx2-xx1])) * (np.max([0, yy2-yy1])) #計(jì)算交集面積
iou = inter_area / (area1+area2-inter_area+1e-6) #計(jì)算交并比
return iou2、GIOU
論文鏈接：Generalized Intersection over Union: A Metric and A Loss for Bounding Box Regression
2.1 為什么提出GIOU
為了解決上面兩個(gè)問題，這篇論文提出了GIOU。由于IoU是比值的概念，對目標(biāo)物體的scale是不敏感的。然而目標(biāo)檢測任務(wù)中的BBox的回歸損失(MSE loss, l1-smooth loss等）優(yōu)化和IoU優(yōu)化不是完全等價(jià)的，而且 Ln 范數(shù)對物體的scale也比較敏感，IoU無法直接優(yōu)化沒有重疊的部分。這篇論文提出可以直接把IoU設(shè)為回歸的loss。
GIOU的計(jì)算很簡單，對于兩個(gè)bounding box A和B。我們可以算出其最小凸集（同時(shí)包含了預(yù)測框和真實(shí)框的最小框的面積）C，有了最小凸集，就可以計(jì)算GIOU，如下所示
從公式可以看出，GIOU有幾個(gè)優(yōu)點(diǎn)：
- GIOU是IOU的下界，且取值范圍為(-1, 1]。當(dāng)兩個(gè)框不重合時(shí)，IOU始終為0，不論A、B相隔多遠(yuǎn)，但是對于GIOU來說，A，B不重合度越高（離的越遠(yuǎn)），GIOU越趨近于-1。
- GIOU其實(shí)就是在IOU的基礎(chǔ)上減掉了一個(gè)東西，這個(gè)減掉的東西，避免了兩個(gè)bbox不重合時(shí)Loss為0的情況;
- 可導(dǎo)：這一點(diǎn)需要強(qiáng)調(diào)下，由于max，min，分段函數(shù)(比如ReLU)這些都是可導(dǎo)的，所以用1-GIOU作為Loss是可導(dǎo)的。
- 與IoU只關(guān)注重疊區(qū)域不同，GIoU不僅關(guān)注重疊區(qū)域，還關(guān)注其他的非重合區(qū)域，能更好的反映兩者的重合度。
但是GIOU同樣存在一些問題，主要有：
- 狀態(tài)1、2、3都是預(yù)測框在目標(biāo)框內(nèi)部且預(yù)測框大小一致的情況，這時(shí)預(yù)測框和目標(biāo)框的差集都是相同的，因此這三種狀態(tài)的GIOU值也都是相同的，這時(shí)GIOU退化成了IOU，無法區(qū)分相對位置關(guān)系；
- GIOU收斂較慢、回歸不夠準(zhǔn)確。
2.2 GIoU代碼
import numpy as np
def Giou_np(bbox_p, bbox_g):
"""
:param bbox_p: predict of bbox(N,4)(x1,y1,x2,y2)
:param bbox_g: groundtruth of bbox(N,4)(x1,y1,x2,y2)
:return:
"""
# for details should go to https://arxiv.org/pdf/1902.09630.pdf
# ensure predict's bbox form
x1p = np.minimum(bbox_p[:, 0], bbox_p[:, 2]).reshape(-1,1)
x2p = np.maximum(bbox_p[:, 0], bbox_p[:, 2]).reshape(-1,1)
y1p = np.minimum(bbox_p[:, 1], bbox_p[:, 3]).reshape(-1,1)
y2p = np.maximum(bbox_p[:, 1], bbox_p[:, 3]).reshape(-1,1)
bbox_p = np.concatenate([x1p, y1p, x2p, y2p], axis=1)
# calc area of Bg
area_p = (bbox_p[:, 2] - bbox_p[:, 0]) * (bbox_p[:, 3] - bbox_p[:, 1])
# calc area of Bp
area_g = (bbox_g[:, 2] - bbox_g[:, 0]) * (bbox_g[:, 3] - bbox_g[:, 1])
# cal intersection
x1I = np.maximum(bbox_p[:, 0], bbox_g[:, 0])
y1I = np.maximum(bbox_p[:, 1], bbox_g[:, 1])
x2I = np.minimum(bbox_p[:, 2], bbox_g[:, 2])
y2I = np.minimum(bbox_p[:, 3], bbox_g[:, 3])
I = np.maximum((y2I - y1I), 0) * np.maximum((x2I - x1I), 0)
# find enclosing box
x1C = np.minimum(bbox_p[:, 0], bbox_g[:, 0])
y1C = np.minimum(bbox_p[:, 1], bbox_g[:, 1])
x2C = np.maximum(bbox_p[:, 2], bbox_g[:, 2])
y2C = np.maximum(bbox_p[:, 3], bbox_g[:, 3])
# calc area of Bc
area_c = (x2C - x1C) * (y2C - y1C)
U = area_p + area_g - I
iou = 1.0 * I / U
# Giou
giou = iou - (area_c - U) / area_c
# loss_iou = 1 - iou loss_giou = 1 - giou
loss_iou = 1.0 - iou
loss_giou = 1.0 - giou
return giou, loss_iou, loss_giou
# def giou_tf
if __name__ == '__main__':
p = np.array([[21,45,103,172],
[34,283,155,406],
[202,174,271,255]])
g = np.array([[59,106,154,230],
[71,272,191,419],
[257,244,329,351]])
Giou_np(p, g)3 DIoU?
論文連接：Distance-IoU Loss: Faster and Better Learning for Bounding Box Regression
3.1 為什么提出DIOU
一個(gè)好的目標(biāo)框回歸函數(shù)應(yīng)該考慮三個(gè)重要幾何因素：重疊面積、中心點(diǎn)距離，長寬比。
針對IOU和GIOU存在的問題，作者從兩個(gè)方面進(jìn)行考慮
- 如何最小化預(yù)測框和目標(biāo)框之間的歸一化距離？
- 如何在預(yù)測框和目標(biāo)框重疊時(shí)，回歸的更準(zhǔn)確？
?針對第一個(gè)問題，提出了DIOU_Loss（Distance_IOU_Loss）
DIOU_Loss考慮了重疊面積和中心點(diǎn)距離，當(dāng)目標(biāo)框包裹預(yù)測框的時(shí)候，直接度量2個(gè)框的距離，因此DIOU_Loss收斂的更快。DIOU損失的優(yōu)點(diǎn)有：
- 與GIoU loss類似，DIoU loss在與目標(biāo)框不重疊時(shí)，仍然可以為邊界框提供移動(dòng)方向。
- DIoU loss可以直接最小化兩個(gè)目標(biāo)框的距離，因此比GIoU loss收斂快得多。
- 對于包含兩個(gè)框在水平方向和垂直方向上這種情況，DIoU損失可以使回歸非常快，而GIoU損失幾乎退化為IoU損失。
- DIoU還可以替換普通的IoU評價(jià)策略，應(yīng)用于NMS中，使得NMS得到的結(jié)果更加合理和有效。
但DIOU同樣存在缺點(diǎn)，那就是沒有考慮到長寬比。比如下面三種情況，目標(biāo)框包裹預(yù)測框，本來DIOU_Loss可以起作用。但預(yù)測框的中心點(diǎn)的位置都是一樣的，因此按照DIOU_Loss的計(jì)算公式，三者的值都是相同的。
3.2 DIOU代碼
def Diou(bboxes1, bboxes2):
rows = bboxes1.shape[0]
cols = bboxes2.shape[0]
dious = torch.zeros((rows, cols))
if rows * cols == 0:#
return dious
exchange = False
if bboxes1.shape[0] > bboxes2.shape[0]:
bboxes1, bboxes2 = bboxes2, bboxes1
dious = torch.zeros((cols, rows))
exchange = True
# #xmin,ymin,xmax,ymax->[:,0],[:,1],[:,2],[:,3]
w1 = bboxes1[:, 2] - bboxes1[:, 0]
h1 = bboxes1[:, 3] - bboxes1[:, 1]
w2 = bboxes2[:, 2] - bboxes2[:, 0]
h2 = bboxes2[:, 3] - bboxes2[:, 1]
area1 = w1 * h1
area2 = w2 * h2
center_x1 = (bboxes1[:, 2] + bboxes1[:, 0]) / 2
center_y1 = (bboxes1[:, 3] + bboxes1[:, 1]) / 2
center_x2 = (bboxes2[:, 2] + bboxes2[:, 0]) / 2
center_y2 = (bboxes2[:, 3] + bboxes2[:, 1]) / 2
inter_max_xy = torch.min(bboxes1[:, 2:],bboxes2[:, 2:])
inter_min_xy = torch.max(bboxes1[:, :2],bboxes2[:, :2])
out_max_xy = torch.max(bboxes1[:, 2:],bboxes2[:, 2:])
out_min_xy = torch.min(bboxes1[:, :2],bboxes2[:, :2])
inter = torch.clamp((inter_max_xy - inter_min_xy), min=0)
inter_area = inter[:, 0] * inter[:, 1]
inter_diag = (center_x2 - center_x1)**2 + (center_y2 - center_y1)**2
outer = torch.clamp((out_max_xy - out_min_xy), min=0)
outer_diag = (outer[:, 0] ** 2) + (outer[:, 1] ** 2)
union = area1+area2-inter_area
dious = inter_area / union - (inter_diag) / outer_diag
dious = torch.clamp(dious,min=-1.0,max = 1.0)
if exchange:
dious = dious.T
return dious4 CIOU
論文鏈接：Distance-IoU Loss: Faster and Better Learning for Bounding Box Regression
4.1 為什么提出CIOU
?CIOU論文考慮到bbox回歸三要素中的長寬比還沒被考慮到計(jì)算中，因此，進(jìn)一步在DIoU的基礎(chǔ)上提出了CIoU。其懲罰項(xiàng)如下面公式：
其中是權(quán)重函數(shù)，而用來度量長寬比的相似性，定義為
?完整的CIOU損失函數(shù)定義：
4.2 CIOU代碼
def bbox_overlaps_ciou(bboxes1, bboxes2):
rows = bboxes1.shape[0]
cols = bboxes2.shape[0]
cious = torch.zeros((rows, cols))
if rows * cols == 0:
return cious
exchange = False
if bboxes1.shape[0] > bboxes2.shape[0]:
bboxes1, bboxes2 = bboxes2, bboxes1
cious = torch.zeros((cols, rows))
exchange = True
w1 = bboxes1[:, 2] - bboxes1[:, 0]
h1 = bboxes1[:, 3] - bboxes1[:, 1]
w2 = bboxes2[:, 2] - bboxes2[:, 0]
h2 = bboxes2[:, 3] - bboxes2[:, 1]
area1 = w1 * h1
area2 = w2 * h2
center_x1 = (bboxes1[:, 2] + bboxes1[:, 0]) / 2
center_y1 = (bboxes1[:, 3] + bboxes1[:, 1]) / 2
center_x2 = (bboxes2[:, 2] + bboxes2[:, 0]) / 2
center_y2 = (bboxes2[:, 3] + bboxes2[:, 1]) / 2
inter_max_xy = torch.min(bboxes1[:, 2:],bboxes2[:, 2:])
inter_min_xy = torch.max(bboxes1[:, :2],bboxes2[:, :2])
out_max_xy = torch.max(bboxes1[:, 2:],bboxes2[:, 2:])
out_min_xy = torch.min(bboxes1[:, :2],bboxes2[:, :2])
inter = torch.clamp((inter_max_xy - inter_min_xy), min=0)
inter_area = inter[:, 0] * inter[:, 1]
inter_diag = (center_x2 - center_x1)**2 + (center_y2 - center_y1)**2
outer = torch.clamp((out_max_xy - out_min_xy), min=0)
outer_diag = (outer[:, 0] ** 2) + (outer[:, 1] ** 2)
union = area1+area2-inter_area
u = (inter_diag) / outer_diag
iou = inter_area / union
with torch.no_grad():
arctan = torch.atan(w2 / h2) - torch.atan(w1 / h1)
v = (4 / (math.pi ** 2)) * torch.pow((torch.atan(w2 / h2) - torch.atan(w1 / h1)), 2)
S = 1 - iou
alpha = v / (S + v)
w_temp = 2 * w1
ar = (8 / (math.pi ** 2)) * arctan * ((w1 - w_temp) * h1)
cious = iou - (u + alpha * ar)
cious = torch.clamp(cious,min=-1.0,max = 1.0)
if exchange:
cious = cious.T
return cious5 EIOU
論文鏈接：Focal and Efficient IOU Loss for Accurate Bounding Box Regression
5.1 為什么提出EIOU?
CIOU Loss雖然考慮了邊界框回歸的重疊面積、中心點(diǎn)距離、縱橫比。但是通過其公式中的v反映的縱橫比的差異，而不是寬高分別與其置信度的真實(shí)差異，所以有時(shí)會(huì)阻礙模型有效的優(yōu)化相似性，于是提出EIOU，它的主要思想是：
- ?一是認(rèn)為CIoU loss對于長寬比加入loss的設(shè)計(jì)不太合理，于是將CIoU loss中反應(yīng)長寬比一致性的部分替換成了分別對于長和寬的一致性loss，形成了EIoU loss。
- ? ? ? ? 二是認(rèn)為不太好的回歸樣本對回歸loss產(chǎn)生了比較大的影響，回歸質(zhì)量相對較好的樣本則難以進(jìn)一步優(yōu)化，所以論文提出Focal EIoU loss進(jìn)行回歸質(zhì)量較好和質(zhì)量較差的樣本之間的平衡。
EIOU Loss優(yōu)點(diǎn)：
- 1）將縱橫比的損失項(xiàng)拆分成預(yù)測的寬高分別與最小外接框?qū)捀叩牟钪?，加速了收斂提高了回歸精度。
- 2）引入了Focal Loss優(yōu)化了邊界框回歸任務(wù)中的樣本不平衡問題，即減少與目標(biāo)框重疊較少的大量錨框?qū)Box 回歸的優(yōu)化貢獻(xiàn)，使回歸過程專注于高質(zhì)量錨框。
5.2 EIOU代碼
def bbox_iou(box1, box2, x1y1x2y2=True, GIoU=False, DIoU=False, CIoU=False, EIoU=False, eps=1e-7):
# Returns the IoU of box1 to box2. box1 is 4, box2 is nx4
box2 = box2.T
# Get the coordinates of bounding boxes
if x1y1x2y2: # x1, y1, x2, y2 = box1
b1_x1, b1_y1, b1_x2, b1_y2 = box1[0], box1[1], box1[2], box1[3]
b2_x1, b2_y1, b2_x2, b2_y2 = box2[0], box2[1], box2[2], box2[3]
else: # transform from xywh to xyxy
b1_x1, b1_x2 = box1[0] - box1[2] / 2, box1[0] + box1[2] / 2
b1_y1, b1_y2 = box1[1] - box1[3] / 2, box1[1] + box1[3] / 2
b2_x1, b2_x2 = box2[0] - box2[2] / 2, box2[0] + box2[2] / 2
b2_y1, b2_y2 = box2[1] - box2[3] / 2, box2[1] + box2[3] / 2
# Intersection area
inter = (torch.min(b1_x2, b2_x2) - torch.max(b1_x1, b2_x1)).clamp(0) * \
(torch.min(b1_y2, b2_y2) - torch.max(b1_y1, b2_y1)).clamp(0)
# Union Area
w1, h1 = b1_x2 - b1_x1, b1_y2 - b1_y1 + eps
w2, h2 = b2_x2 - b2_x1, b2_y2 - b2_y1 + eps
union = w1 * h1 + w2 * h2 - inter + eps
iou = inter / union
if GIoU or DIoU or CIoU or EIoU:
cw = torch.max(b1_x2, b2_x2) - torch.min(b1_x1, b2_x1) # convex (smallest enclosing box) width
ch = torch.max(b1_y2, b2_y2) - torch.min(b1_y1, b2_y1) # convex height
if CIoU or DIoU or EIoU: # Distance or Complete IoU https://arxiv.org/abs/1911.08287v1
c2 = cw ** 2 + ch ** 2 + eps # convex diagonal squared
rho2 = ((b2_x1 + b2_x2 - b1_x1 - b1_x2) ** 2 +
(b2_y1 + b2_y2 - b1_y1 - b1_y2) ** 2) / 4 # center distance squared
if DIoU:
return iou - rho2 / c2 # DIoU
elif CIoU: # https://github.com/Zzh-tju/DIoU-SSD-pytorch/blob/master/utils/box/box_utils.py#L47
v = (4 / math.pi ** 2) * torch.pow(torch.atan(w2 / h2) - torch.atan(w1 / h1), 2)
with torch.no_grad():
alpha = v / (v - iou + (1 + eps))
return iou - (rho2 / c2 + v * alpha) # CIoU
elif EIoU:
rho_w2 = ((b2_x2 - b2_x1) - (b1_x2 - b1_x1)) ** 2
rho_h2 = ((b2_y2 - b2_y1) - (b1_y2 - b1_y1)) ** 2
cw2 = cw ** 2 + eps
ch2 = ch ** 2 + eps
return iou - (rho2 / c2 + rho_w2 / cw2 + rho_h2 / ch2)
else: # GIoU https://arxiv.org/pdf/1902.09630.pdf
c_area = cw * ch + eps # convex area
return iou - (c_area - union) / c_area # GIoU
else:
return iou # IoU
6?Wise-IoU
論文鏈接：Wise-IoU: Bounding Box Regression Loss with Dynamic Focusing Mechanism
具體關(guān)于Wise-IoU損失的介紹請參考前期博客
優(yōu)化改進(jìn)YOLOv5算法之Wise-IOU損失函數(shù)_yolov5算法優(yōu)化_AI追隨者的博客-CSDN博客
7 YOLOv5中添加GIoU、DIoU、CIoU、EIoU、Wise-IoU損失函數(shù)
?yolov5-6.1版本中的iou損失函數(shù)是在utils/metrics.py文件定義的，在該文件添加以下關(guān)于GIoU、DIoU、CIoU、EIoU、Wise-IoU函數(shù)的代碼，如下所示
import numpy as np
import torch, math
class WIoU_Scale:
''' monotonous: {
None: origin v1
True: monotonic FM v2
False: non-monotonic FM v3
}
momentum: The momentum of running mean'''
iou_mean = 1.
monotonous = False
_momentum = 1 - 0.5 ** (1 / 7000)
_is_train = True
def __init__(self, iou):
self.iou = iou
self._update(self)
@classmethod
def _update(cls, self):
if cls._is_train: cls.iou_mean = (1 - cls._momentum) * cls.iou_mean + \
cls._momentum * self.iou.detach().mean().item()
@classmethod
def _scaled_loss(cls, self, gamma=1.9, delta=3):
if isinstance(self.monotonous, bool):
if self.monotonous:
return (self.iou.detach() / self.iou_mean).sqrt()
else:
beta = self.iou.detach() / self.iou_mean
alpha = delta * torch.pow(gamma, beta - delta)
return beta / alpha
return 1
def bbox_iou(box1, box2, xywh=True, GIoU=False, DIoU=False, CIoU=False, SIoU=False, EIoU=False, WIoU=False, Focal=False, alpha=1, gamma=0.5, scale=False, eps=1e-7):
# Returns Intersection over Union (IoU) of box1(1,4) to box2(n,4)
# Get the coordinates of bounding boxes
if xywh: # transform from xywh to xyxy
(x1, y1, w1, h1), (x2, y2, w2, h2) = box1.chunk(4, -1), box2.chunk(4, -1)
w1_, h1_, w2_, h2_ = w1 / 2, h1 / 2, w2 / 2, h2 / 2
b1_x1, b1_x2, b1_y1, b1_y2 = x1 - w1_, x1 + w1_, y1 - h1_, y1 + h1_
b2_x1, b2_x2, b2_y1, b2_y2 = x2 - w2_, x2 + w2_, y2 - h2_, y2 + h2_
else: # x1, y1, x2, y2 = box1
b1_x1, b1_y1, b1_x2, b1_y2 = box1.chunk(4, -1)
b2_x1, b2_y1, b2_x2, b2_y2 = box2.chunk(4, -1)
w1, h1 = b1_x2 - b1_x1, (b1_y2 - b1_y1).clamp(eps)
w2, h2 = b2_x2 - b2_x1, (b2_y2 - b2_y1).clamp(eps)
# Intersection area
inter = (b1_x2.minimum(b2_x2) - b1_x1.maximum(b2_x1)).clamp(0) * \
(b1_y2.minimum(b2_y2) - b1_y1.maximum(b2_y1)).clamp(0)
# Union Area
union = w1 * h1 + w2 * h2 - inter + eps
if scale:
self = WIoU_Scale(1 - (inter / union))
# IoU
# iou = inter / union # ori iou
iou = torch.pow(inter/(union + eps), alpha) # alpha iou
if CIoU or DIoU or GIoU or EIoU or SIoU or WIoU:
cw = b1_x2.maximum(b2_x2) - b1_x1.minimum(b2_x1) # convex (smallest enclosing box) width
ch = b1_y2.maximum(b2_y2) - b1_y1.minimum(b2_y1) # convex height
if CIoU or DIoU or EIoU or SIoU or WIoU: # Distance or Complete IoU https://arxiv.org/abs/1911.08287v1
c2 = (cw ** 2 + ch ** 2) ** alpha + eps # convex diagonal squared
rho2 = (((b2_x1 + b2_x2 - b1_x1 - b1_x2) ** 2 + (b2_y1 + b2_y2 - b1_y1 - b1_y2) ** 2) / 4) ** alpha # center dist ** 2
if CIoU: # https://github.com/Zzh-tju/DIoU-SSD-pytorch/blob/master/utils/box/box_utils.py#L47
v = (4 / math.pi ** 2) * (torch.atan(w2 / h2) - torch.atan(w1 / h1)).pow(2)
with torch.no_grad():
alpha_ciou = v / (v - iou + (1 + eps))
if Focal:
return iou - (rho2 / c2 + torch.pow(v * alpha_ciou + eps, alpha)), torch.pow(inter/(union + eps), gamma) # Focal_CIoU
else:
return iou - (rho2 / c2 + torch.pow(v * alpha_ciou + eps, alpha)) # CIoU
elif EIoU:
rho_w2 = ((b2_x2 - b2_x1) - (b1_x2 - b1_x1)) ** 2
rho_h2 = ((b2_y2 - b2_y1) - (b1_y2 - b1_y1)) ** 2
cw2 = torch.pow(cw ** 2 + eps, alpha)
ch2 = torch.pow(ch ** 2 + eps, alpha)
if Focal:
return iou - (rho2 / c2 + rho_w2 / cw2 + rho_h2 / ch2), torch.pow(inter/(union + eps), gamma) # Focal_EIou
else:
return iou - (rho2 / c2 + rho_w2 / cw2 + rho_h2 / ch2) # EIou
elif SIoU:
# SIoU Loss https://arxiv.org/pdf/2205.12740.pdf
s_cw = (b2_x1 + b2_x2 - b1_x1 - b1_x2) * 0.5 + eps
s_ch = (b2_y1 + b2_y2 - b1_y1 - b1_y2) * 0.5 + eps
sigma = torch.pow(s_cw ** 2 + s_ch ** 2, 0.5)
sin_alpha_1 = torch.abs(s_cw) / sigma
sin_alpha_2 = torch.abs(s_ch) / sigma
threshold = pow(2, 0.5) / 2
sin_alpha = torch.where(sin_alpha_1 > threshold, sin_alpha_2, sin_alpha_1)
angle_cost = torch.cos(torch.arcsin(sin_alpha) * 2 - math.pi / 2)
rho_x = (s_cw / cw) ** 2
rho_y = (s_ch / ch) ** 2
gamma = angle_cost - 2
distance_cost = 2 - torch.exp(gamma * rho_x) - torch.exp(gamma * rho_y)
omiga_w = torch.abs(w1 - w2) / torch.max(w1, w2)
omiga_h = torch.abs(h1 - h2) / torch.max(h1, h2)
shape_cost = torch.pow(1 - torch.exp(-1 * omiga_w), 4) + torch.pow(1 - torch.exp(-1 * omiga_h), 4)
if Focal:
return iou - torch.pow(0.5 * (distance_cost + shape_cost) + eps, alpha), torch.pow(inter/(union + eps), gamma) # Focal_SIou
else:
return iou - torch.pow(0.5 * (distance_cost + shape_cost) + eps, alpha) # SIou
elif WIoU:
if Focal:
raise RuntimeError("WIoU do not support Focal.")
elif scale:
return getattr(WIoU_Scale, '_scaled_loss')(self), (1 - iou) * torch.exp((rho2 / c2)), iou # WIoU https://arxiv.org/abs/2301.10051
else:
return iou, torch.exp((rho2 / c2)) # WIoU v1
if Focal:
return iou - rho2 / c2, torch.pow(inter/(union + eps), gamma) # Focal_DIoU
else:
return iou - rho2 / c2 # DIoU
c_area = cw * ch + eps # convex area
if Focal:
return iou - torch.pow((c_area - union) / c_area + eps, alpha), torch.pow(inter/(union + eps), gamma) # Focal_GIoU https://arxiv.org/pdf/1902.09630.pdf
else:
return iou - torch.pow((c_area - union) / c_area + eps, alpha) # GIoU https://arxiv.org/pdf/1902.09630.pdf
if Focal:
return iou, torch.pow(inter/(union + eps), gamma) # Focal_IoU
else:
return iou # IoU然后在utils/loss.py文件中調(diào)用bbox_iou損失函數(shù)時(shí)，將對應(yīng)的IOU設(shè)置為True即可。?
參考文章：【深度學(xué)習(xí)小知識(shí)】目標(biāo)檢測中的IOU、GIOU、DIOU、CIOU、EIOU等理論解析_你好?。海┑牟┛?CSDN博客?深入淺出Yolo系列之Yolov3&Yolov4&Yolov5&Yolox核心基礎(chǔ)知識(shí)完整講解 - 知乎文章來源：http://www.zghlxwxcb.cn/news/detail-443691.html
詳解IoU、GIoU、DIoU、CIoU、EIoU和DIoU-NMS_小Aer的博客-CSDN博客文章來源地址http://www.zghlxwxcb.cn/news/detail-443691.html
到了這里，關(guān)于優(yōu)化改進(jìn)YOLOv5算法之添加GIoU、DIoU、CIoU、EIoU、Wise-IoU模塊(超詳細(xì))的文章就介紹完了。如果您還想了解更多內(nèi)容，請?jiān)谟疑辖撬阉鱐OY模板網(wǎng)以前的文章或繼續(xù)瀏覽下面的相關(guān)文章，希望大家以后多多支持TOY模板網(wǎng)！
