# Copyright (C) 2024 Intel Corporation
# SPDX-License-Identifier: Apache-2.0
#
"""HungarianMatcher3D module for 3d object detection."""
import numpy as np
import torch
from scipy.optimize import linear_sum_assignment
from torch import nn
from otx.algo.common.utils.bbox_overlaps import bbox_overlaps
from otx.algo.object_detection_3d.utils.utils import box_cxcylrtb_to_xyxy
[docs]
class HungarianMatcher3D(nn.Module):
"""This class computes an assignment between the targets and the predictions of the network."""
def __init__(
self,
cost_class: float = 1.0,
cost_3dcenter: float = 1.0,
cost_bbox: float = 1.0,
cost_giou: float = 1.0,
):
"""Creates the matcher.
Args:
cost_class (float): This is the relative weight of the classification error in the matching cost.
cost_3dcenter (float): This is the relative weight of the L1 error of the 3d center in the matching cost.
cost_bbox (float): This is the relative weight of the L1 error of the bbox coordinates in the matching cost.
cost_giou (float): This is the relative weight of the giou loss of the bbox in the matching cost.
"""
super().__init__()
self.cost_class = cost_class
self.cost_3dcenter = cost_3dcenter
self.cost_bbox = cost_bbox
self.cost_giou = cost_giou
[docs]
@torch.no_grad()
def forward(self, outputs: dict, targets: list, group_num: int = 11) -> list:
"""Performs the matching.
Args:
outputs: This is a dict that contains at least these entries:
"scores": Tensor of dim [batch_size, num_queries, num_classes] with the classification logits
"boxes_3d": Tensor of dim [batch_size, num_queries, 4] with the predicted 3d box coordinates
targets: This is a list of targets (len(targets) = batch_size), where each target is a dict containing:
"labels": Tensor of dim [num_target_boxes] (where num_target_boxes is the number of ground-truth
objects in the target) containing the class labels
"boxes": Tensor of dim [num_target_boxes, 4] containing the target box coordinates
Returns:
A list of size batch_size, containing tuples of (index_i, index_j) where:
- index_i is the indices of the selected predictions (in order)
- index_j is the indices of the corresponding selected targets (in order)
For each batch element, it holds:
len(index_i) = len(index_j) = min(num_queries, num_target_boxes)
"""
bs, num_queries = outputs["boxes_3d"].shape[:2]
# We flatten to compute the cost matrices in a batch
out_prob = outputs["scores"].flatten(0, 1).sigmoid() # [batch_size * num_queries, num_classes]
# Also concat the target labels and boxes
tgt_ids = torch.cat([v["labels"] for v in targets]).long()
# Compute the classification cost.
alpha = 0.25
gamma = 2.0
neg_cost_class = (1 - alpha) * (out_prob**gamma) * (-(1 - out_prob + 1e-8).log())
pos_cost_class = alpha * ((1 - out_prob) ** gamma) * (-(out_prob + 1e-8).log())
cost_class = pos_cost_class[:, tgt_ids] - neg_cost_class[:, tgt_ids]
out_3dcenter = outputs["boxes_3d"][:, :, 0:2].flatten(0, 1) # [batch_size * num_queries, 4]
tgt_3dcenter = torch.cat([v["boxes_3d"][:, 0:2] for v in targets])
# Compute the 3dcenter cost between boxes
cost_3dcenter = torch.cdist(out_3dcenter, tgt_3dcenter, p=1)
out_2dbbox = outputs["boxes_3d"][:, :, 2:6].flatten(0, 1) # [batch_size * num_queries, 4]
tgt_2dbbox = torch.cat([v["boxes_3d"][:, 2:6] for v in targets])
# Compute the L1 cost between boxes
cost_bbox = torch.cdist(out_2dbbox, tgt_2dbbox, p=1)
# Compute the giou cost betwen boxes
out_bbox = outputs["boxes_3d"].flatten(0, 1) # [batch_size * num_queries, 4]
tgt_bbox = torch.cat([v["boxes_3d"] for v in targets])
cost_giou = -bbox_overlaps(
box_cxcylrtb_to_xyxy(out_bbox),
box_cxcylrtb_to_xyxy(tgt_bbox),
mode="giou",
)
# Final cost matrix
c = (
self.cost_bbox * cost_bbox
+ self.cost_3dcenter * cost_3dcenter
+ self.cost_class * cost_class
+ self.cost_giou * cost_giou
)
c = c.view(bs, num_queries, -1).cpu()
sizes = [len(v["boxes"]) for v in targets]
# indices = [linear_sum_assignment(c[i]) for i, c in enumerate(C.split(sizes, -1))]
indices = []
g_num_queries = num_queries // group_num
c_list = c.split(g_num_queries, dim=1)
for g_i in range(group_num):
c_g = c_list[g_i]
indices_g = [linear_sum_assignment(c[i]) for i, c in enumerate(c_g.split(sizes, -1))]
if g_i == 0:
indices = indices_g
else:
indices = [
(
np.concatenate([indice1[0], indice2[0] + g_num_queries * g_i]),
np.concatenate([indice1[1], indice2[1]]),
)
for indice1, indice2 in zip(indices, indices_g)
]
return [(torch.as_tensor(i, dtype=torch.int64), torch.as_tensor(j, dtype=torch.int64)) for i, j in indices]