|
@@ -170,7 +170,7 @@ class InstanceSegPredictor(DetPredictor):
|
|
|
box_idx_start = 0
|
|
box_idx_start = 0
|
|
|
pred_box = []
|
|
pred_box = []
|
|
|
|
|
|
|
|
- if isinstance(pred, list) and len(pred[0]) == 4:
|
|
|
|
|
|
|
+ if isinstance(pred[0], list) and len(pred[0]) == 4:
|
|
|
# Adapt to SOLOv2, which only support prediction with a batch_size of 1.
|
|
# Adapt to SOLOv2, which only support prediction with a batch_size of 1.
|
|
|
pred_class_id = [[pred_[1], pred_[2]] for pred_ in pred]
|
|
pred_class_id = [[pred_[1], pred_[2]] for pred_ in pred]
|
|
|
pred_mask = [pred_[3] for pred_ in pred]
|
|
pred_mask = [pred_[3] for pred_ in pred]
|
|
@@ -181,7 +181,7 @@ class InstanceSegPredictor(DetPredictor):
|
|
|
}
|
|
}
|
|
|
for i in range(len(pred_class_id))
|
|
for i in range(len(pred_class_id))
|
|
|
]
|
|
]
|
|
|
- if isinstance(pred, list) and len(pred[0]) == 3:
|
|
|
|
|
|
|
+ if isinstance(pred[0], list) and len(pred[0]) == 3:
|
|
|
# Adapt to PP-YOLOE_seg-S, which only support prediction with a batch_size of 1.
|
|
# Adapt to PP-YOLOE_seg-S, which only support prediction with a batch_size of 1.
|
|
|
return [
|
|
return [
|
|
|
{"boxes": np.array(pred[i][0]), "masks": np.array(pred[i][2])}
|
|
{"boxes": np.array(pred[i][0]), "masks": np.array(pred[i][2])}
|