论文地址:https://arxiv.org/pdf/2110.13389.pdf
目录
前言:
wasserstein_loss
LOSS.PY
最近在做目标检测时,图片分辨率为6016 x 2048,遇到一个200*200的小目标问题,这里加入了NWP针对该问题做出优化
utils/metrics.py中加入wasserstein_loss
def wasserstein_loss(pred, target, eps=1e-7, constant=12.8):
"""Implementation of paper `A Normalized Gaussian Wasserstein Distance for
Tiny Object Detection .
Args:
pred (Tensor): Predicted bboxes of format (cx, cy, w, h),
shape (n, 4).
target (Tensor): Corresponding gt bboxes, shape (n, 4).
eps (float): Eps to avoid log(0).
Return:
Tensor: Loss tensor.
"""
center1 = pred[:, :2]
center2 = target[:, :2]
whs = center1[:, :2] - center2[:, :2]
center_distance = whs[:, 0] * whs[:, 0] + whs[:, 1] * whs[:, 1] + eps
w1 = pred[:, 2] + eps
h1 = pred[:, 3] + eps
w2 = target[:, 2] + eps
h2 = target[:, 3] + eps
wh_distance = ((w1 - w2) ** 2 + (h1 - h2) ** 2) / 4
wasserstein_2 = center_distance + wh_distance
return torch.exp(-torch.sqrt(wasserstein_2) / constant)
修改loss.py
if n:
# pxy, pwh, _, pcls = pi[b, a, gj, gi].tensor_split((2, 4, 5), dim=1) # faster, requires torch 1.8.0
pxy, pwh, _, pcls = pi[b, a, gj, gi].split((2, 2, 1, self.nc), 1) # target-subset of predictions
# 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 修改
# Objectness
# score_iou = iou.detach().clamp(0).type(tobj.dtype) #修改
nwd = wasserstein_loss(pbox, tbox[i]).squeeze()
nwd_ratio = 0.5 # 平衡稀疏 nwd和 iou各取0.5 如果数据集全是小目标的换可以设置1或者0.9,,08 意思为只用nwd
lbox += (1 - nwd_ratio) * (1.0 - nwd).mean() + nwd_ratio * (1.0 - iou).mean()
# Objectness
iou = (iou.detach() * nwd_ratio + nwd.detach() * (1 - nwd_ratio)).clamp(0, 1).type(tobj.dtype) #这里clamp(0,1)中设置了最小参数必须大于等于0进行
# ===============================================
# Objectness
iou = iou.detach().clamp(0).type(tobj.dtype)
结语:
后续更新对小目标改进:有效BiFormer注意力机制