def test_default_anchor_generator_centered(self):
cfg = BaseDetectionConfig()
cfg.MODEL.ANCHOR_GENERATOR.SIZES = [[32, 64]]
cfg.MODEL.ANCHOR_GENERATOR.ASPECT_RATIOS = [[0.25, 1, 4]]
cfg.MODEL.ANCHOR_GENERATOR.OFFSET = 0.5
anchor_generator = DefaultAnchorGenerator(cfg, [ShapeSpec(stride=4)])
# only the last two dimensions of features matter here
num_images = 2
features = {"stage3": torch.rand(num_images, 96, 1, 2)}
anchors = anchor_generator([features["stage3"]])
expected_anchor_tensor = torch.tensor(
[
[-30.0, -6.0, 34.0, 10.0],
[-14.0, -14.0, 18.0, 18.0],
[-6.0, -30.0, 10.0, 34.0],
[-62.0, -14.0, 66.0, 18.0],
[-30.0, -30.0, 34.0, 34.0],
[-14.0, -62.0, 18.0, 66.0],
[-26.0, -6.0, 38.0, 10.0],
[-10.0, -14.0, 22.0, 18.0],
[-2.0, -30.0, 14.0, 34.0],
[-58.0, -14.0, 70.0, 18.0],
[-26.0, -30.0, 38.0, 34.0],
[-10.0, -62.0, 22.0, 66.0],
]
)
for i in range(num_images):
assert torch.allclose(anchors[i][0].tensor, expected_anchor_tensor)
def __init__(self, cfg, input_shape: Dict[str, ShapeSpec]):
super().__init__()
# fmt: off
self.min_box_side_len = cfg.MODEL.PROPOSAL_GENERATOR.MIN_SIZE
self.in_features = cfg.MODEL.RPN.IN_FEATURES
self.nms_thresh = cfg.MODEL.RPN.NMS_THRESH
self.nms_type = cfg.MODEL.RPN.NMS_TYPE
self.batch_size_per_image = cfg.MODEL.RPN.BATCH_SIZE_PER_IMAGE
self.positive_fraction = cfg.MODEL.RPN.POSITIVE_FRACTION
self.smooth_l1_beta = cfg.MODEL.RPN.SMOOTH_L1_BETA
self.loss_weight = cfg.MODEL.RPN.LOSS_WEIGHT
# fmt: on
# Map from self.training state to train/test settings
self.pre_nms_topk = {
True: cfg.MODEL.RPN.PRE_NMS_TOPK_TRAIN,
False: cfg.MODEL.RPN.PRE_NMS_TOPK_TEST,
}
self.post_nms_topk = {
True: cfg.MODEL.RPN.POST_NMS_TOPK_TRAIN,
False: cfg.MODEL.RPN.POST_NMS_TOPK_TEST,
}
self.boundary_threshold = cfg.MODEL.RPN.BOUNDARY_THRESH
self.anchor_generator = DefaultAnchorGenerator(
cfg, [input_shape[f] for f in self.in_features])
self.box2box_transform = Box2BoxTransform(
weights=cfg.MODEL.RPN.BBOX_REG_WEIGHTS)
self.anchor_matcher = Matcher(cfg.MODEL.RPN.IOU_THRESHOLDS,
cfg.MODEL.RPN.IOU_LABELS,
allow_low_quality_matches=True)
self.rpn_head = StandardRPNHead(
cfg, [input_shape[f] for f in self.in_features])
def test_default_anchor_generator(self):
cfg = BaseDetectionConfig()
cfg.MODEL.ANCHOR_GENERATOR.SIZES = [[32, 64]]
cfg.MODEL.ANCHOR_GENERATOR.ASPECT_RATIOS = [[0.25, 1, 4]]
anchor_generator = DefaultAnchorGenerator(cfg, [ShapeSpec(stride=4)])
# only the last two dimensions of features matter here
num_images = 2
features = {"stage3": torch.rand(num_images, 96, 1, 2)}
anchors = anchor_generator([features["stage3"]])
expected_anchor_tensor = torch.tensor(
[
[-32.0, -8.0, 32.0, 8.0],
[-16.0, -16.0, 16.0, 16.0],
[-8.0, -32.0, 8.0, 32.0],
[-64.0, -16.0, 64.0, 16.0],
[-32.0, -32.0, 32.0, 32.0],
[-16.0, -64.0, 16.0, 64.0],
[-28.0, -8.0, 36.0, 8.0], # -28.0 == -32.0 + STRIDE (4)
[-12.0, -16.0, 20.0, 16.0],
[-4.0, -32.0, 12.0, 32.0],
[-60.0, -16.0, 68.0, 16.0],
[-28.0, -32.0, 36.0, 32.0],
[-12.0, -64.0, 20.0, 64.0],
]
)
for i in range(num_images):
assert torch.allclose(anchors[i][0].tensor, expected_anchor_tensor)
def __init__(self, cfg, input_shape: List[ShapeSpec]):
super().__init__()
# Standard RPN is shared across levels:
in_channels = [s.channels for s in input_shape]
assert len(
set(in_channels)) == 1, "Each level must have the same channel!"
in_channels = in_channels[0]
# RPNHead should take the same input as anchor generator
# NOTE: it assumes that creating an anchor generator does not have unwanted side effect.
anchor_generator = DefaultAnchorGenerator(cfg, input_shape)
num_cell_anchors = anchor_generator.num_cell_anchors
box_dim = anchor_generator.box_dim
assert (len(set(num_cell_anchors)) == 1
), "Each level must have the same number of cell anchors"
num_cell_anchors = num_cell_anchors[0]
# 3x3 conv for the hidden representation
self.conv = nn.Conv2d(in_channels,
in_channels,
kernel_size=3,
stride=1,
padding=1)
# 1x1 conv for predicting objectness logits
self.objectness_logits = nn.Conv2d(in_channels,
num_cell_anchors,
kernel_size=1,
stride=1)
# 1x1 conv for predicting box2box transform deltas
self.anchor_deltas = nn.Conv2d(in_channels,
num_cell_anchors * box_dim,
kernel_size=1,
stride=1)
for l in [self.conv, self.objectness_logits, self.anchor_deltas]:
nn.init.normal_(l.weight, std=0.01)
nn.init.constant_(l.bias, 0)
def build_anchor_generator(cfg, input_shape):
return DefaultAnchorGenerator(cfg, input_shape)
