Ng2dZddlZddlmZddlmZddlmZddlm Z ddl m Z dd lm Z dd l mZd ZGd d eZdS)aBase target assigner module. The job of a TargetAssigner is, for a given set of anchors (bounding boxes) and groundtruth detections (bounding boxes), to assign classification and regression targets to each anchor as well as weights to each anchor (specifying, e.g., which anchors should not contribute to training loss). It assigns classification/regression targets by performing the following steps: 1) Computing pairwise similarity between anchors and groundtruth boxes using a provided RegionSimilarity Calculator 2) Computing a matching based on the similarity matrix using a provided Matcher 3) Assigning regression targets based on the matching and a provided BoxCoder 4) Assigning classification targets based on the matching and groundtruth labels Note that TargetAssigners only operate on detections from a single image at a time, so any logic for applying a TargetAssigner to multiple images must be handled externally. N)Optional)box_list) IouSimilarity) ArgMaxMatcher)Match)BoxList)FasterRcnnBoxCoder keypointsc eZdZdZddefdedededede ed e f d Z dd e d e fd Z d e d e defdZdejfdZdefdZdefdZdefdZdZdS)TargetAssignerzATarget assigner to compute classification and regression targets.g?Nsimilarity_calcmatcher box_codernegative_class_weightunmatched_cls_targetkeypoints_field_namecn||_||_||_||_|||_nd|_||_dS)ajConstruct Object Detection Target Assigner. Args: similarity_calc: a RegionSimilarityCalculator matcher: Matcher used to match groundtruth to anchors. box_coder: BoxCoder used to encode matching groundtruth boxes with respect to anchors. negative_class_weight: classification weight to be associated to negative anchors (default: 1.0). The weight must be in [0., 1.]. unmatched_cls_target: a float32 tensor with shape [d_1, d_2, ..., d_k] which is consistent with the classification target for each anchor (and can be empty for scalar targets). This shape must thus be compatible with the groundtruth labels that are passed to the "assign" function (which have shape [num_gt_boxes, d_1, d_2, ..., d_k]). If set to None, unmatched_cls_target is set to be [0] for each anchor. Raises: ValueError: if similarity_calc is not a RegionSimilarityCalculator or if matcher is not a Matcher or if box_coder is not a BoxCoder N)_similarity_calc_matcher _box_coder_negative_class_weight_unmatched_cls_target_keypoints_field_name)selfrrrrrrs c/var/www/html/ai-engine/env/lib/python3.11/site-packages/effdet/object_detection/target_assigner.py__init__zTargetAssigner.__init__2sI4!0 #&;#  +)=D & &)+D &%9"""anchorsgroundtruth_boxesctt|tjstdt|tjstd|j||}|j|}||||}| ||}|||fS)aAssign classification and regression targets to each anchor. For a given set of anchors and groundtruth detections, match anchors to groundtruth_boxes and assign classification and regression targets to each anchor as well as weights based on the resulting match (specifying, e.g., which anchors should not contribute to training loss). Anchors that are not matched to anything are given a classification target of self._unmatched_cls_target which can be specified via the constructor. Args: anchors: a BoxList representing N anchors groundtruth_boxes: a BoxList representing M groundtruth boxes groundtruth_labels: a tensor of shape [M, d_1, ... d_k] with labels for each of the ground_truth boxes. The subshape [d_1, ... d_k] can be empty (corresponding to scalar inputs). When set to None, groundtruth_labels assumes a binary problem where all ground_truth boxes get a positive label (of 1). groundtruth_weights: a float tensor of shape [M] indicating the weight to assign to all anchors match to a particular groundtruth box. The weights must be in [0., 1.]. If None, all weights are set to 1. **params: Additional keyword arguments for specific implementations of the Matcher. Returns: cls_targets: a float32 tensor with shape [num_anchors, d_1, d_2 ... d_k], where the subshape [d_1, ..., d_k] is compatible with groundtruth_labels which has shape [num_gt_boxes, d_1, d_2, ... d_k]. cls_weights: a float32 tensor with shape [num_anchors] reg_targets: a float32 tensor with shape [num_anchors, box_code_dimension] reg_weights: a float32 tensor with shape [num_anchors] match: a matcher.Match object encoding the match between anchors and groundtruth boxes, with rows corresponding to groundtruth boxes and columns corresponding to anchors. Raises: ValueError: if anchors or groundtruth_boxes are not of type box_list.BoxList zanchors must be an BoxListz$groundtruth_boxes must be an BoxList) isinstancerr ValueErrorrcomparermatch_create_regression_targets_create_classification_targets) rr r!groundtruth_labelsgroundtruth_weightsmatch_quality_matrixr& reg_targets cls_targetss rassignzTargetAssigner.assignVsZ'8#344 ;9:: :+X-=>> ECDD D $4<<=NPWXX ##$89955g?PRWXX 99:LeTT K..rr&c|}tjd|}||||}t j|}||jrs| |j}tjd|j ddz|} ||| | } | |j| |j ||} |||jj dd} |} tj| d| | }|S)aOReturns a regression target for each anchor. Args: anchors: a BoxList representing N anchors groundtruth_boxes: a BoxList representing M groundtruth_boxes match: a matcher.Match object Returns: reg_targets: a float32 tensor with shape [N, box_code_dimension] )rdeviceunmatched_value ignored_value)rrNr)r2torchzerosgather_based_on_matchboxesrr has_fieldr get_fieldshape add_fieldrencode_default_regression_targetrepeat match_resultsmatched_column_indicatorwhere unsqueeze)rr r!r&r2zero_boxmatched_gt_boxesmatched_gt_boxlistgroundtruth_keypointszero_kpmatched_keypointsmatched_reg_targetsunmatched_ignored_reg_targetsmatched_anchors_maskr,s rr'z)TargetAssigner._create_regression_targetss!!;vf555 66  # # % %xx7YY%-.>??  & &t'A B B X$5$?$?@Z$[$[ !k$)>)DQRR)H"HQWXXXG % ; ;%wg!<!W!W   ( ()CEV W W W"o445GQQ(,(G(G(O(O(V(VW\WjWpqrWsuv(w(w%$==??k"6"@"@"C"CEXZwxx rr2c^tjd|j|S)aReturns the default target for anchors to regress to. Default regression targets are set to zero (though in this implementation what these targets are set to should not matter as the regression weight of any box set to regress to the default target is zero). Returns: default_target: a float32 tensor with shape [1, box_code_dimension] rr1)r6r7r code_size)rr2s rr?z)TargetAssigner._default_regression_targets){1do7799&IIIIrcF|||j|jS)a|Create classification targets for each anchor. Assign a classification target of for each anchor to the matching groundtruth label that is provided by match. Anchors that are not matched to anything are given the target self._unmatched_cls_target Args: groundtruth_labels: a tensor of shape [num_gt_boxes, d_1, ... d_k] with labels for each of the ground_truth boxes. The subshape [d_1, ... d_k] can be empty (corresponding to scalar labels). match: a matcher.Match object that provides a matching between anchors and groundtruth boxes. Returns: a float32 tensor with shape [num_anchors, d_1, d_2 ... d_k], where the subshape [d_1, ..., d_k] is compatible with groundtruth_labels which has shape [num_gt_boxes, d_1, d_2, ... d_k]. r3)r8r)rr)r&s rr(z-TargetAssigner._create_classification_targetss2&**  6dF`+bb brc2||ddS)aqSet regression weight for each anchor. Only positive anchors are set to contribute to the regression loss, so this method returns a weight of 1 for every positive anchor and 0 for every negative anchor. Args: match: a matcher.Match object that provides a matching between anchors and groundtruth boxes. groundtruth_weights: a float tensor of shape [M] indicating the weight to assign to all anchors match to a particular groundtruth box. Returns: a float32 tensor with shape [num_anchors] representing regression weights. rr5r4)r8rr&r*s r_create_regression_weightsz)TargetAssigner._create_regression_weightss!**+>bbd*eeerc<||d|jS)arCreate classification weights for each anchor. Positive (matched) anchors are associated with a weight of positive_class_weight and negative (unmatched) anchors are associated with a weight of negative_class_weight. When anchors are ignored, weights are set to zero. By default, both positive/negative weights are set to 1.0, but they can be adjusted to handle class imbalance (which is almost always the case in object detection). Args: match: a matcher.Match object that provides a matching between anchors and groundtruth boxes. groundtruth_weights: a float tensor of shape [M] indicating the weight to assign to all anchors match to a particular groundtruth box. Returns: a float32 tensor with shape [num_anchors] representing classification weights. rrR)r8rrSs r_create_classification_weightsz-TargetAssigner._create_classification_weightss.$** r4C^+`` `rc|jS)z\Get BoxCoder of this TargetAssigner. Returns: BoxCoder object. )r)rs rrzTargetAssigner.box_coders r)NN)__name__ __module__ __qualname____doc__KEYPOINTS_FIELD_NAMErrr floatrstrrr r.rr'r6r2r?r(rTrVrrrr r /s]KK14]a-A":": ": ":Zl":(-":KSTY?":'*":":":":HC/C/gC/'C/C/C/C/J'g^c@ J J J J Jbbbbb.fffff"`E````*rr )r[r6typingrrregion_similarity_calculatorrargmax_matcherrrrr rr r\objectr r_rrres$ 777777))))))))))))"[[[[[V[[[[[r