def compute_box_size_targets(hw_map: np.ndarray, boxes: np.ndarray) -> np.ndarray:
y_center, x_center = np_frame_ops.boxes_clipped_centers(boxes)
heights, widths = np_frame_ops.boxes_heights_widths(boxes)
fm_height, fm_width = hw_map.shape[:2]
for k, (y, x) in enumerate(zip(y_center, x_center)):
yi = int(y * fm_height)
xi = int(x * fm_width)
hw_map[yi, xi, 0] = heights[k] * fm_height
hw_map[yi, xi, 1] = widths[k] * fm_width
hw_map[yi, xi, 2] = 1
return hw_map
def draw_labels_frame_boxes(
image: Union[np.ndarray, tf.Tensor],
boxes: Union[np.ndarray, tf.Tensor],
labels: Union[np.ndarray, tf.Tensor],
figsize: Tuple[int, int] = (6, 6),
title: Optional[str] = None,
num_classes: Optional[int] = None,
fontsize: int = 10,
linewidth: float = 2,
fmt: str = "png",
) -> Image:
if isinstance(image, tf.Tensor):
image = image.numpy()
if isinstance(boxes, tf.Tensor):
boxes = boxes.numpy()
if isinstance(labels, tf.Tensor):
labels = labels.numpy()
if num_classes is None:
num_classes = np.max(labels)
num_boxes = boxes.shape[0]
label2color = labels_to_colors(num_classes)
figure = plt.figure(figsize=figsize)
ax = figure.add_subplot(111, aspect="equal")
if title is not None:
plt.title(title)
if image.dtype != np.uint8 and image.max() > 2:
image = image / 255.0
plt.imshow(image)
height, width = image.shape[:2]
heights, widths = boxes_heights_widths(boxes.astype(np.float32))
y_min, x_min = boxes[:, 0], boxes[:, 1]
for idx in range(num_boxes):
pos = (int(x_min[idx] * width), int(y_min[idx] * height))
label = labels[idx]
color = label2color.get(label, "r")
plt.text(*pos, f"{label}", color=color, size=fontsize)
rect = patches.Rectangle(
pos,
int(widths[idx] * height),
int(heights[idx] * width),
linewidth=linewidth,
edgecolor=color,
facecolor="none",
)
ax.add_patch(rect)
return plot_to_image(figure, format=fmt)
def from_tf_boxes(
boxes: np.ndarray,
labels: np.ndarray,
scores: Optional[np.ndarray] = None,
classes_scores: Optional[np.ndarray] = None,
) -> "BoxDetectionOutput":
y, c = np_frame_ops.boxes_centers(boxes)
h, w = np_frame_ops.boxes_heights_widths(boxes)
boxes = np.stack([h, w, y, c], axis=-1)
return BoxDetectionOutput(
boxes=boxes,
labels=labels,
scores=scores or np.ones(shape=labels.shape),
classes_scores=classes_scores or np.ones(shape=labels.shape),
)
def compute_mean_box_size_targets(hw_map: np.ndarray, boxes: np.ndarray) -> np.ndarray:
y_center, x_center = np_frame_ops.boxes_clipped_centers(boxes)
heights, widths = np_frame_ops.boxes_heights_widths(boxes)
fm_height, fm_width = hw_map.shape[:2]
for k, (y, x) in enumerate(zip(y_center, x_center)):
yi = int(y * fm_height)
xi = int(x * fm_width)
hw_map[yi, xi, 0] += heights[k] * fm_height
hw_map[yi, xi, 1] += widths[k] * fm_width
hw_map[yi, xi, 2] += 1
hw_map[:, :, :2] = hw_map[:, :, :2] / (hw_map[:, :, 2:] + 1e-6)
# make weights in range (0, 1)
hw_map[:, :, 2] = np.minimum(hw_map[:, :, 2], 1)
return hw_map