MCPcopy Create free account
hub / github.com/OpenGVLab/HumanBench / masks_to_boxes

Function masks_to_boxes

PATH/core/models/ops/box_ops.py:92–116  ·  view source on GitHub ↗

Compute the bounding boxes around the provided masks The masks should be in format [N, H, W] where N is the number of masks, (H, W) are the spatial dimensions. Returns a [N, 4] tensors, with the boxes in xyxy format

(masks)

Source from the content-addressed store, hash-verified

90
91
92def masks_to_boxes(masks):
93 """Compute the bounding boxes around the provided masks
94
95 The masks should be in format [N, H, W] where N is the number of masks, (H, W) are the spatial dimensions.
96
97 Returns a [N, 4] tensors, with the boxes in xyxy format
98 """
99 if masks.numel() == 0:
100 return torch.zeros((0, 4), device=masks.device)
101
102 h, w = masks.shape[-2:]
103
104 y = torch.arange(0, h, dtype=torch.float)
105 x = torch.arange(0, w, dtype=torch.float)
106 y, x = torch.meshgrid(y, x)
107
108 x_mask = (masks * x.unsqueeze(0))
109 x_max = x_mask.flatten(1).max(-1)[0]
110 x_min = x_mask.masked_fill(~(masks.bool()), 1e8).flatten(1).min(-1)[0]
111
112 y_mask = (masks * y.unsqueeze(0))
113 y_max = y_mask.flatten(1).max(-1)[0]
114 y_min = y_mask.masked_fill(~(masks.bool()), 1e8).flatten(1).min(-1)[0]
115
116 return torch.stack([x_min, y_min, x_max, y_max], 1)

Callers

nothing calls this directly

Calls 1

stackMethod · 0.80

Tested by

no test coverage detected