FPhY>ddlZddlmZddlmcmZddlmZmZddl m Z ddl m Z GddejZGdd eZy) N) FocalLoss VarifocalLoss)bbox_iou)HungarianMatcherceZdZdZ d fd Zd dZd dZ d dZedZ dZ d dZ d d Z xZ S)DETRLossa+ DETR (DEtection TRansformer) Loss class. This class calculates and returns the different loss components for the DETR object detection model. It computes classification loss, bounding box loss, GIoU loss, and optionally auxiliary losses. Attributes: nc (int): The number of classes. loss_gain (dict): Coefficients for different loss components. aux_loss (bool): Whether to compute auxiliary losses. use_fl (bool): Use FocalLoss or not. use_vfl (bool): Use VarifocalLoss or not. use_uni_match (bool): Whether to use a fixed layer to assign labels for the auxiliary branch. uni_match_ind (int): The fixed indices of a layer to use if `use_uni_match` is True. matcher (HungarianMatcher): Object to compute matching cost and indices. fl (FocalLoss or None): Focal Loss object if `use_fl` is True, otherwise None. vfl (VarifocalLoss or None): Varifocal Loss object if `use_vfl` is True, otherwise None. device (torch.device): Device on which tensors are stored. ct|| ddddddd}||_tdddd|_||_||_|r tnd|_|r tnd|_ ||_ ||_ d|_ y) a DETR loss function. Args: nc (int): The number of classes. loss_gain (dict): The coefficient of loss. aux_loss (bool): If 'aux_loss = True', loss at each decoder layer are to be used. use_vfl (bool): Use VarifocalLoss or not. use_uni_match (bool): Whether to use a fixed layer to assign labels for auxiliary branch. uni_match_ind (int): The fixed indices of a layer. Nrg?)classbboxgiou no_objectmaskdice)r rr) cost_gain)super__init__ncrmatcher loss_gainaux_lossrflrvfl use_uni_match uni_match_inddevice) selfrrruse_fluse_vflrr __class__s hC:\Users\daisl\Desktop\realtime-object-detection\venv\Lib\site-packages\ultralytics/models/utils/loss.pyrzDETRLoss.__init__!s&   "#QUV`abI'AqRS2TU "  !')+T&-=?4** cd|}|jdd\}}tj|||jdzftj|j } | j d|jdd| dddf} |j||d| z}|jrU|r |jr|j||| } n |j|| j} | t|d|z z} n:tjd ||jdj!} || j#|j$d ziS) z^Computes the classification loss based on predictions, target values, and ground truth scores. loss_classNr r)dtyper.none reductionr )shapetorchzerosrint64rscatter_ unsqueezeviewrrfloatmaxnnBCEWithLogitsLossmeansumsqueezer) r pred_scorestargets gt_scoresnum_gtspostfix name_classbsnqone_hotloss_clss r#_get_loss_classzDETRLoss._get_loss_classCs2"'+ ""2A&B++r2tww{35;;w~~^G--b115#ss(#NN2r1-7 7748888KGD77; @ GQ", ,H=r++f=k9UZZ[\]aacHH,,.1HHIIr$cFd|}d|}i}t|dk(rJtjd|j||<tjd|j||<|S|jdt j ||dzt|z ||<d t||d d z ||<||jt|z ||<|jd ||z||<|jDcic]\}}||jc}}Scc}}w) z~Calculates and returns the bounding box loss and GIoU loss for the predicted and ground truth bounding boxes. loss_bbox loss_giourrrr8r*g?T)xywhGIoUr) lenr-tensorrrFl1_lossrr8itemsr9) r pred_bboxes gt_bboxesr> name_bbox name_gioulosskvs r#_get_loss_bboxzDETRLoss._get_loss_bboxYs  y) y)  y>Q #ll2dkkBDO#ll2dkkBDOK..0199[)_d3eehkluhvvYidQU VVYy/--/#i.@Y..04 ?BY+/::<8<41a199;<888s>Dc tj|dnd|j} |O|jrC|j ||j ||j ||||||j nd| }t t||D]d\} \} } ||| nd}|j| | ||||| || }| dxx|d|z cc<| d xx|d |z cc<| d xx|d |z cc<fd || dd|| d d|| d i} | S)zGet auxiliary losses.Nr rImasksgt_mask)r\r]r> match_indicesrr&rrFr rGloss_class_aux loss_bbox_aux loss_giou_aux) r-r.rrrr enumeratezip _get_loss)rrQr:rRgt_cls gt_groupsr^r>r\r]rUi aux_bboxes aux_scores aux_masksloss_s r# _get_loss_auxzDETRLoss._get_loss_auxsp{{ 11qASAST  T%7%7 LLT5G5G)H)4T5G5G)H)2)/)2LQL]uT5G5G/Hcg18 ):M,5Sk5R+S 'A' J$)$5a4INN:#-#,#)#,)2+2+21>#@E Guz'34 4G Guy 23 3G Guy 23 3G,T(WI &QG9 %tAwG9 %tAw0 r$c htjt|Dcgc]\}\}}tj|| c}}}}tj|Dcgc]\}}| c}}}tj|Dcgc]\}}| c}}}||f|fScc}}}wcc}}wcc}}w)z[Returns batch indices, source indices, and destination indices from provided match indices.)r-catrb full_like)r^rgsrc_ batch_idxsrc_idxdstdst_idxs r# _get_indexzDETRLoss._get_indexsII)TaJbcJb;1hsAusA6Jbcd ))?Xc1S?@))?XaS?@7#W,,d??s#B!  B(  B. c tjt||Dcgc]J\}\}}t|dkDr||n.tjd|j d|j Lc}}}}tjt||Dcgc]J\}\}}t|dkDr||n.tjd|j d|j Lc}}}} || fScc}}}wcc}}}w)z[Assigns predicted bounding boxes to ground truth bounding boxes based on the match indices.rr(rI)r-rnrcrLr.r,r) rrQrRr^tIrq pred_assignedJ gt_assigneds r#_get_assigned_bboxeszDETRLoss._get_assigned_bboxess  m<#>< 6AqFQJAaDEKK1772;t{{$S S<#>? ii M:!<: 6AqFQJAaDEKK1772;t{{$S S:!<= k)) #>!r^idxgt_idxr@rAr;r<rUs r#rdzDETRLoss._get_losss7   LL)4)2)/)2/418 ):Moom4 V!,S!19V3DY ""2A&B**b"Xtww{7I7IQWQ]Q]^f~ KKR1C1CD y>%k&8&8&:IDQYYZ\]IcN D((gy#i.Zabc D'' YHI r$c |j|_|jdd}|d|d|d} }}|j|d|d||| ||} |jr,| j |j |dd|dd||| ||| S)a Args: pred_bboxes (torch.Tensor): [l, b, query, 4] pred_scores (torch.Tensor): [l, b, query, num_classes] batch (dict): A dict includes: gt_cls (torch.Tensor) with shape [num_gts, ], gt_bboxes (torch.Tensor): [num_gts, 4], gt_groups (List(int)): a list of batch size length includes the number of gts of each image. postfix (str): postfix of loss name. r^Nclsbboxesrfr(r>r^)rgetrdrrrl) rrQr:batchr>kwargsr^rerRrf total_losss r#forwardzDETRLoss.forwards"((  ?D9 ',U|U8_eKFX9 ^^KO$/O$-$*$-,32? $A  ==   "";s#3["5EyRXZcer#*, -r$)PNTTFFr))NrNN)NNrN)__name__ __module__ __qualname____doc__rrDrXrl staticmethodrvr}rdr __classcell__r"s@r#r r sw($  DJ,9r%)  "/b--*  $$Lr$r c4eZdZdZdfd ZedZxZS)RTDETRDetectionLossa# Real-Time DeepTracker (RT-DETR) Detection Loss class that extends the DETRLoss. This class computes the detection loss for the RT-DETR model, which includes the standard detection loss as well as an additional denoising training loss when provided with denoising metadata. c |\}}t||||}|c|d|d} } t|dt| k(sJ|j| | |d} t||||d| } |j | |S|j |j D cic]'} | dt jd|j)c} |Scc} w)aF Forward pass to compute the detection loss. Args: preds (tuple): Predicted bounding boxes and scores. batch (dict): Batch data containing ground truth information. dn_bboxes (torch.Tensor, optional): Denoising bounding boxes. Default is None. dn_scores (torch.Tensor, optional): Denoising scores. Default is None. dn_meta (dict, optional): Metadata for denoising. Default is None. Returns: (dict): Dictionary containing the total loss and, if applicable, the denoising loss. dn_pos_idx dn_num_grouprf_dnrrHrI) rrrLget_dn_match_indicesrkeysr-rMr)rpredsr dn_bboxes dn_scoresdn_metarQr:rrrr^dn_lossrVr"s r#rzRTDETRDetectionLoss.forwards$) [W_[+uE   '.|'U Ju[)*c*o= ==!55j,PUVaPbcMgoiE5`monG   g &    XbXgXgXijXiST!Cy%,,r$++*NNXij kks,Cchg}tjdg|ddjd}t|D]\}}|dkDrtj|tj ||z}|j |}t||t|k(sJddt||dt|d|j|||f|jtjdgtj tjdgtj f|S) a Get the match indices for denoising. Args: dn_pos_idx (List[torch.Tensor]): List of tensors containing positive indices for denoising. dn_num_group (int): Number of denoising groups. gt_groups (List[int]): List of integers representing the number of ground truths for each image. Returns: (List[tuple]): List of tuples containing matched indices for denoising. rNr()endr'zExpected the same length, zbut got z and z respectively.)r') r- as_tensorcumsum_rbarangelongrepeatrLappendr.)rrrfdn_match_indices idx_groupsrgnum_gtrs r#rz(RTDETRDetectionLoss.get_dn_match_indices>s__a%9)CR.%9:BB1E "9-IAvz& CjQRmS|4:a=)S[8V:VV83z!}-.eCK=O ''A(?@ ''aS )KU[[Z[Y\didndnMo(pq. r$)NNN)rrrrrrrrrs@r#rrs" D  r$r)r-torch.nnr5torch.nn.functional functionalrNultralytics.utils.lossrrultralytics.utils.metricsropsrModuler rr$r#rs= ;.!DryyDNB (B r$