def densepose_chart_predictions_to_dict(instances):
segmentations = ToMaskConverter.convert(instances.pred_densepose,
instances.pred_boxes,
instances.image_size)
results = []
for k in range(len(instances)):
densepose_results_quantized = quantize_densepose_chart_result(
ToChartResultConverter.convert(instances.pred_densepose[k],
instances.pred_boxes[k]))
densepose_results_quantized.labels_uv_uint8 = (
densepose_results_quantized.labels_uv_uint8.cpu())
segmentation = segmentations.tensor[k]
segmentation_encoded = mask_utils.encode(
np.require(segmentation.numpy(),
dtype=np.uint8,
requirements=["F"]))
segmentation_encoded["counts"] = segmentation_encoded["counts"].decode(
"utf-8")
result = {
"densepose": densepose_results_quantized,
"segmentation": segmentation_encoded,
}
results.append(result)
return results
def __call__(self, instances: Instances) -> BitMasks:
"""
Converts predicted data from `instances` into the GT mask data
Args:
instances (Instances): predicted results, expected to have `pred_densepose` field
Returns:
Boolean Tensor of the size of the input image that has non-zero
values at pixels that are estimated to belong to the detected object
"""
return ToMaskConverter.convert(instances.pred_densepose,
instances.pred_boxes,
instances.image_size)