def __init__(self, cfg):
# self.cls_loss_func = SigmoidFocalLoss(
# cfg.MODEL.LOC.LOSS_GAMMA,
# cfg.MODEL.LOC.LOSS_ALPHA
# )
cls_loss_name = cfg.MODEL.LOC.CLS_LOSS
self.cls_divide_pos_num = True
if cls_loss_name == 'fixed_focal_loss':
self.cls_loss_func = FixSigmoidFocalLoss(cfg.MODEL.LOC.LOSS_GAMMA,
cfg.MODEL.LOC.LOSS_ALPHA)
elif cls_loss_name == 'L2':
self.cls_loss_func = L2LossWithLogit()
elif cls_loss_name == 'GHMC':
self.cls_loss_func = GHMC(
bins=cfg.MODEL.LOC.LOSS_GHMC_BINS,
alpha=cfg.MODEL.LOC.LOSS_GHMC_ALPHA,
momentum=cfg.MODEL.LOC.LOSS_GHMC_MOMENTUM)
self.cls_divide_pos_num = False
# we make use of IOU Loss for bounding boxes regression,
# but we found that L1 in log scale can yield a similar performance
self.box_reg_loss_func = IOULoss()
if cfg.MODEL.LOC.TARGET_GENERATOR == 'fcos' and cfg.MODEL.LOC.FCOS_CENTERNESS:
self.centerness_loss_func = nn.BCEWithLogitsLoss()
self.prepare_targets = build_target_generator(cfg)
self.cls_loss_weight = cfg.MODEL.LOC.CLS_WEIGHT
self.centerness_weight_reg = cfg.MODEL.LOC.TARGET_GENERATOR == 'fcos' and cfg.MODEL.LOC.FCOS_CENTERNESS_WEIGHT_REG
self.debug_vis_labels = cfg.MODEL.LOC.DEBUG.VIS_LABELS
self.cls_divide_pos_sum = cfg.MODEL.LOC.DIVIDE_POS_SUM
if self.cls_divide_pos_sum:
self.cls_divide_pos_num = False
def __init__(self, cfg):
self.cls_loss_func = SigmoidFocalLoss(cfg.MODEL.FCOS.LOSS_GAMMA,
cfg.MODEL.FCOS.LOSS_ALPHA)
# we make use of IOU Loss for bounding boxes regression,
# but we found that L1 in log scale can yield a similar performance
self.box_reg_loss_func = IOULoss()
self.centerness_loss_func = nn.BCEWithLogitsLoss()
def __init__(self, gamma, alpha, stripe_width=32):
self.cls_loss_func = SigmoidFocalLoss(gamma, alpha)
# we make use of IOU Loss for bounding boxes regression,
# but we found that L1 in log scale can yield a similar performance
self.box_reg_loss_func = IOULoss()
self.centerness_loss_func = nn.BCEWithLogitsLoss()
self.stripe_width = stripe_width
def __init__(self, cfg):
self.cls_loss_func = SigmoidFocalLoss([cfg.MODEL.AVOD.LOSS_GAMMA],
[cfg.MODEL.AVOD.LOSS_ALPHA])
# we make use of IOU Loss for bounding boxes regression,
# but we found that L1 in log scale can yield a similar performance
self.box_reg_loss_func = IOULoss()
self.point_reg_loss_func = nn.L1Loss()
self.confidance_loss_func = nn.BCEWithLogitsLoss()
def __init__(self, cfg):
self.cls_loss_func = SigmoidFocalLoss(cfg.MODEL.FCOS.LOSS_GAMMA,
cfg.MODEL.FCOS.LOSS_ALPHA)
# we make use of IOU Loss for bounding boxes regression,
# but we found that L1 in log scale can yield a similar performance
self.box_reg_loss_func = IOULoss()
self.centerness_loss_func = nn.BCEWithLogitsLoss()
self.vis_labels = cfg.MODEL.FCOS.DEBUG.VIS_LABELS
self.object_sizes_of_interest = cfg.MODEL.FCOS.OBJECT_SIZES
def __init__(self, cfg):
self.cls_loss_func = SigmoidFocalLoss(cfg.MODEL.FCOS.LOSS_GAMMA,
cfg.MODEL.FCOS.LOSS_ALPHA)
self.fpn_strides = cfg.MODEL.FCOS.FPN_STRIDES
self.center_sampling_radius = cfg.MODEL.FCOS.CENTER_SAMPLING_RADIUS
self.iou_loss_type = cfg.MODEL.FCOS.IOU_LOSS_TYPE
self.norm_reg_targets = cfg.MODEL.FCOS.NORM_REG_TARGETS
# we make use of IOU Loss for bounding boxes regression,
# but we found that L1 in log scale can yield a similar performance
self.box_reg_loss_func = IOULoss(self.iou_loss_type)
self.centerness_loss_func = nn.BCEWithLogitsLoss(reduction="sum")
def __init__(self, cfg):
self.cls_loss_func = SigmoidFocalLoss(cfg.MODEL.FCOS.LOSS_GAMMA,
cfg.MODEL.FCOS.LOSS_ALPHA)
self.center_sample = cfg.MODEL.FCOS.CENTER_SAMPLE
self.strides = cfg.MODEL.FCOS.FPN_STRIDES
self.radius = cfg.MODEL.FCOS.POS_RADIUS
self.loc_loss_type = cfg.MODEL.FCOS.LOC_LOSS_TYPE
# we make use of IOU Loss for bounding boxes regression,
# but we found that L1 in log scale can yield a similar performance
self.box_reg_loss_func = IOULoss(self.loc_loss_type)
self.centerness_loss_func = nn.BCEWithLogitsLoss()
self.dense_points = cfg.MODEL.FCOS.DENSE_POINTS
def __init__(self, cfg):
self.cls_loss_func = SigmoidFocalLoss(cfg.MODEL.FCOS.LOSS_GAMMA,
cfg.MODEL.FCOS.LOSS_ALPHA)
# we make use of IOU Loss for bounding boxes regression,
# but we found that L1 in log scale can yield a similar performance
self.box_reg_loss_func = IOULoss()
self.centerness_loss_func = nn.BCEWithLogitsLoss()
self.cascade_area_th = cfg.MODEL.FCOS.CASCADE_AREA_TH
self.vis_labels = cfg.MODEL.FCOS.DEBUG.VIS_LABELS
self.no_match_gt_count = {
pos_area: 0
for pos_area in self.cascade_area_th
}
def __init__(self, cfg):
self.cls_loss_func = SigmoidFocalLoss(cfg.MODEL.FCOS.LOSS_GAMMA,
cfg.MODEL.FCOS.LOSS_ALPHA)
# we make use of IOU Loss for bounding boxes regression,
# but we found that L1 in log scale can yield a similar performance
self.box_reg_loss_func = IOULoss()
self.centerness_loss_func = nn.BCEWithLogitsLoss()
# generate sizes of interest
soi = []
prev_size = -1
for s in cfg.MODEL.FCOS.SIZES_OF_INTEREST:
soi.append([prev_size, s])
prev_size = s
soi.append([prev_size, INF])
self.object_sizes_of_interest = soi
def __init__(self, cfg):
if cfg.LOSS.FOCAL_LOSS == 'SIGMOID':
self.cls_loss_func = SigmoidFocalLoss(cfg.MODEL.FCOS.LOSS_GAMMA,
cfg.MODEL.FCOS.LOSS_ALPHA)
elif cfg.LOSS.FOCAL_LOSS == 'SOFTMAX':
self.cls_loss_func = SoftmaxFocalLoss(cfg.MODEL.FCOS.LOSS_GAMMA,
cfg.MODEL.FCOS.LOSS_ALPHA)
else:
raise Exception('no focal loss type named {}'.format(
cfg.LOSS.FOCAL_LOSS))
self.center_sample = cfg.MODEL.FCOS.CENTER_SAMPLE
self.strides = cfg.MODEL.FCOS.FPN_STRIDES
self.radius = cfg.MODEL.FCOS.POS_RADIUS
self.loc_loss_type = cfg.MODEL.FCOS.LOC_LOSS_TYPE
# we make use of IOU Loss for bounding boxes regression,
# but we found that L1 in log scale can yield a similar performance
self.box_reg_loss_func = IOULoss(self.loc_loss_type)
self.centerness_loss_func = nn.BCEWithLogitsLoss()
self.dense_points = cfg.MODEL.FCOS.DENSE_POINTS
def __init__(self, cfg):
self.cls_loss_func = SigmoidFocalLoss(cfg.MODEL.FCOS.LOSS_GAMMA,
cfg.MODEL.FCOS.LOSS_ALPHA)
self.use_p2 = cfg.MODEL.BACKBONE.USE_P2
# pdb.set_trace()
# (Pdb) cfg.MODEL.FCOS.LOSS_GAMMA
# 2.0
# (Pdb) cfg.MODEL.FCOS.LOSS_ALPHA
# 0.25
self.center_sample = cfg.MODEL.FCOS.CENTER_SAMPLE
if self.use_p2:
self.strides = cfg.MODEL.FCOS.FPN_STRIDES_ADDP2
else:
self.strides = cfg.MODEL.FCOS.FPN_STRIDES
self.radius = cfg.MODEL.FCOS.POS_RADIUS
self.loc_loss_type = cfg.MODEL.FCOS.LOC_LOSS_TYPE
# we make use of IOU Loss for bounding boxes regression,
# but we found that L1 in log scale can yield a similar performance
self.box_reg_loss_func = IOULoss(self.loc_loss_type)
self.centerness_loss_func = nn.BCEWithLogitsLoss()
def __init__(self, cfg):
# we make use of IOU Loss for bounding boxes regression,
# but we found that L1 in log scale can yield a similar performance
self.box_reg_loss_func = IOULoss()
self.scale = cfg.MODEL.SCALE
