hQSSKrSSKJr SSKJr SSKJrJr SSKJ r SSK J r "SS\R5r "S S \ 5rSS jrSS jrS rSSjrg)N)LOGGER)bbox_iouprobiou)xywhr2xyxyxyxy) TORCH_1_11c ^\rSrSrSrSS\S\S\S\S\4 U4Sjjjr\R"5S 5r S r S r S r S rSSjrSr\SSj5r\S5rSrU=r$)TaskAlignedAssigner aG A task-aligned assigner for object detection. This class assigns ground-truth (gt) objects to anchors based on the task-aligned metric, which combines both classification and localization information. Attributes: topk (int): The number of top candidates to consider. num_classes (int): The number of object classes. alpha (float): The alpha parameter for the classification component of the task-aligned metric. beta (float): The beta parameter for the localization component of the task-aligned metric. eps (float): A small value to prevent division by zero. topk num_classesalphabetaepsc^>[TU]5 XlX lX0lX@lXPlg)a Initialize a TaskAlignedAssigner object with customizable hyperparameters. Args: topk (int, optional): The number of top candidates to consider. num_classes (int, optional): The number of object classes. alpha (float, optional): The alpha parameter for the classification component of the task-aligned metric. beta (float, optional): The beta parameter for the localization component of the task-aligned metric. eps (float, optional): A small value to prevent division by zero. N)super__init__r r rrr)selfr r rrr __class__s O/home/james-whalen/.local/lib/python3.13/site-packages/ultralytics/utils/tal.pyrTaskAlignedAssigner.__init__s*  &  c^ URSUlURSUlURm URS:Xa[R "USUR 5[R"U5[R"U5[R"US5[R"US54$URXX4XV5$![RRae [R"S5 XX4XV4Vs/sHowR5PM Os snfnnUR"U6n [U 4SjU 55s$f=f)a Compute the task-aligned assignment. Args: pd_scores (torch.Tensor): Predicted classification scores with shape (bs, num_total_anchors, num_classes). pd_bboxes (torch.Tensor): Predicted bounding boxes with shape (bs, num_total_anchors, 4). anc_points (torch.Tensor): Anchor points with shape (num_total_anchors, 2). gt_labels (torch.Tensor): Ground truth labels with shape (bs, n_max_boxes, 1). gt_bboxes (torch.Tensor): Ground truth boxes with shape (bs, n_max_boxes, 4). mask_gt (torch.Tensor): Mask for valid ground truth boxes with shape (bs, n_max_boxes, 1). Returns: target_labels (torch.Tensor): Target labels with shape (bs, num_total_anchors). target_bboxes (torch.Tensor): Target bounding boxes with shape (bs, num_total_anchors, 4). target_scores (torch.Tensor): Target scores with shape (bs, num_total_anchors, num_classes). fg_mask (torch.Tensor): Foreground mask with shape (bs, num_total_anchors). target_gt_idx (torch.Tensor): Target ground truth indices with shape (bs, num_total_anchors). References: https://github.com/Nioolek/PPYOLOE_pytorch/blob/master/ppyoloe/assigner/tal_assigner.py rr).rz7CUDA OutOfMemoryError in TaskAlignedAssigner, using CPUc3D># UHoRT5v M g7fN)to).0tdevices r .TaskAlignedAssigner.forward..Xs6v!fvs )shapebs n_max_boxesrtorch full_liker zeros_like_forwardcudaOutOfMemoryErrorrwarningcputuple) r pd_scores pd_bboxes anc_points gt_labels gt_bboxesmask_gtr cpu_tensorsresultrs @rforwardTaskAlignedAssigner.forward-s&.//!$$??1-!!   q  & 143C3CD  +  +  6!23  6!23   7==zia azz** 7 NNT U-6:Zc,mn,mq557,mnKn]]K0F6v66 6  7sC;ED.-,EEcvURXXEX65upxn URXyUR5upnURXEX5upnX-nUR SSS9nX-R SSS9nUU-XR -- R S5R S5nUU-nXXR5U 4$)a Compute the task-aligned assignment. Args: pd_scores (torch.Tensor): Predicted classification scores with shape (bs, num_total_anchors, num_classes). pd_bboxes (torch.Tensor): Predicted bounding boxes with shape (bs, num_total_anchors, 4). anc_points (torch.Tensor): Anchor points with shape (num_total_anchors, 2). gt_labels (torch.Tensor): Ground truth labels with shape (bs, n_max_boxes, 1). gt_bboxes (torch.Tensor): Ground truth boxes with shape (bs, n_max_boxes, 4). mask_gt (torch.Tensor): Mask for valid ground truth boxes with shape (bs, n_max_boxes, 1). Returns: target_labels (torch.Tensor): Target labels with shape (bs, num_total_anchors). target_bboxes (torch.Tensor): Target bounding boxes with shape (bs, num_total_anchors, 4). target_scores (torch.Tensor): Target scores with shape (bs, num_total_anchors, num_classes). fg_mask (torch.Tensor): Foreground mask with shape (bs, num_total_anchors). target_gt_idx (torch.Tensor): Target ground truth indices with shape (bs, num_total_anchors). T)dimkeepdim) get_pos_maskselect_highest_overlapsr$ get_targetsamaxr unsqueezebool)rr.r/r0r1r2r3mask_pos align_metricoverlaps target_gt_idxfg_mask target_labels target_bboxes target_scorespos_align_metrics pos_overlapsnorm_align_metrics rr(TaskAlignedAssigner._forwardZs&,0+<+< ) , (,0+G+G\`\l\l+m( 7;6F6Fy]j6t3 m  (--"d-C +11b$1G )L8ioo^g^n^nokk477D,<,ITTU^_I z1-{{7== 9K[K[\ yArr2%**YM]M]^tyy!A  % %b)Aq!!e)O*Dd K" !!,"2A6w}}--rc[R"UR[RURS9SnX5UR --nUR 5R5UnURSURS5UnURS5 [R"URSURSUR4[RURS9nURSURS5S5 USS2SS2S4RSSUR5n [R "U S:US5nXgU4$) a' Compute target labels, target bounding boxes, and target scores for the positive anchor points. Args: gt_labels (torch.Tensor): Ground truth labels of shape (b, max_num_obj, 1), where b is the batch size and max_num_obj is the maximum number of objects. gt_bboxes (torch.Tensor): Ground truth bounding boxes of shape (b, max_num_obj, 4). target_gt_idx (torch.Tensor): Indices of the assigned ground truth objects for positive anchor points, with shape (b, h*w), where h*w is the total number of anchor points. fg_mask (torch.Tensor): A boolean tensor of shape (b, h*w) indicating the positive (foreground) anchor points. Returns: target_labels (torch.Tensor): Target labels for positive anchor points with shape (b, h*w). target_bboxes (torch.Tensor): Target bounding boxes for positive anchor points with shape (b, h*w, 4). target_scores (torch.Tensor): Target scores for positive anchor points with shape (b, h*w, num_classes). )r\rZr).Nr9rrrYr[N)r%r_r#int64rr$r^flattenr`r"rmr]r scatter_rArepeatwhere) rr1r2rFrG batch_indrHrIrJfg_scores_masks rr?TaskAlignedAssigner.get_targetss3(LLTWWEKK HXHXYZcd %D4D4D(DD !(002=A "r9??2+>? N  Q   #]%8%8%;T=M=M N++ '' q-"9"9""=qA At,33Aq$:J:JK NQ$6 qI ]::rc.URSnURupEnURSSS5RSS5upx[R"USU- XS- 4SS9RXEUS5n U R S5R U5$) a  Select positive anchor centers within ground truth bounding boxes. Args: xy_centers (torch.Tensor): Anchor center coordinates, shape (h*w, 2). gt_bboxes (torch.Tensor): Ground truth bounding boxes, shape (b, n_boxes, 4). eps (float, optional): Small value for numerical stability. Returns: (torch.Tensor): Boolean mask of positive anchors, shape (b, n_boxes, h*w). Note: b: batch size, n_boxes: number of ground truth boxes, h: height, w: width. Bounding box format: [x_min, y_min, x_max, y_max]. rr9rr[Nr:)r"r`chunkr%catamingt_) xy_centersr2r n_anchorsr#n_boxes_ltrb bbox_deltass rrQ,TaskAlignedAssigner.select_candidates_in_gtss"$$Q' "QAq)//15iiD!1B!6=M8M NTUV[[\^irtvw "&&s++rcURS5nUR5S:aURS5S:RSUS5nUR S5n[ R "URURURS9nURSURS5S5 [ R"XFU5R5nURS5nUR S5nXsU4$)a Select anchor boxes with highest IoU when assigned to multiple ground truths. Args: mask_pos (torch.Tensor): Positive mask, shape (b, n_max_boxes, h*w). overlaps (torch.Tensor): IoU overlaps, shape (b, n_max_boxes, h*w). n_max_boxes (int): Maximum number of ground truth boxes. Returns: target_gt_idx (torch.Tensor): Indices of assigned ground truths, shape (b, h*w). fg_mask (torch.Tensor): Foreground mask, shape (b, h*w). mask_pos (torch.Tensor): Updated positive mask, shape (b, n_max_boxes, h*w). r<rr9rY) sumrrrArTargmaxr%r]r"rZrrrfloat)rCrEr$rGmask_multi_gtsmax_overlaps_idxis_max_overlapsrFs rr>+TaskAlignedAssigner.select_highest_overlaps+s ,,r" ;;=1 %//2Q6>>r;PRSN'q1 #kk(..W_WfWfgO  $ $Q(8(B(B1(Eq I{{>HMSSUHll2&G + x//r)rrr#rr$r r ) Pg?g@& .>r)r)__name__ __module__ __qualname____firstlineno____doc__intrrr%no_gradr6r(r=rRrbrSr? staticmethodrQr>__static_attributes__ __classcell__)rs@rr r s SCUX]jo$ ]]_*7*7X#ZJ06 &D [.@);V,,,00rr c.\rSrSrSrSr\S5rSrg)RotatedTaskAlignedAssigneriJzSAssigns ground-truth objects to rotated bounding boxes using a task-aligned metric.cT[X5RS5RS5$)z)Calculate IoU for rotated bounding boxes.r9r)rrarmrns rrb*RotatedTaskAlignedAssigner.iou_calculationMs#y,44R8??BBrc[U5nURSSS9up4pVXC- nXc- nX- n Xw-RSS9n X-RSS9n X-RSS9n X-RSS9n U S:X:*-U S:-X:*-$)aM Select the positive anchor center in gt for rotated bounding boxes. Args: xy_centers (torch.Tensor): Anchor center coordinates with shape (h*w, 2). gt_bboxes (torch.Tensor): Ground truth bounding boxes with shape (b, n_boxes, 5). Returns: (torch.Tensor): Boolean mask of positive anchors with shape (b, n_boxes, h*w). rr<rr9r)rsplitr)rr2cornersabrdabadapnorm_abnorm_ad ap_dot_ab ap_dot_ads rrQ3RotatedTaskAlignedAssigner.select_candidates_in_gtsQs!+]]1"]- a U U^7--B-'7--B-'WMMbM) WMMbM) Q9#78INKyOcddrN) rrrrrrbrrQrrrrrrJs!]Ceerrc //pCUceUSRUSRpe[U5GHupx[U[5(aXR SSO[ XS5[ XS54up[R"XUS9U-n [R"XUS9U-n [(a[R"XSS9O[R"X5upUR[R"X4S5RSS55 UR[R"X-S4XUS 95 GM [R"U5[R"U54$) zGenerate anchors from features.Nrr[r)r\rrZij)indexingr9rY)rZr enumerate isinstancelistr"rr%r_rmeshgridappendstackr`fullr) featsstridesgrid_cell_offset anchor_points stride_tensorrZristridehwsxsys r make_anchorsrms7#%r=   !HNNE!HOO6w' %/t%<%rsQ &#{0")){0| e!4 eF > (a -r