Add ComputeLoss() class (#1950)

test.py

@@ -13,7 +13,6 @@ from models.experimental import attempt_load
 from utils.datasets import create_dataloader
 from utils.general import coco80_to_coco91_class, check_dataset, check_file, check_img_size, check_requirements, \
     box_iou, non_max_suppression, scale_coords, xyxy2xywh, xywh2xyxy, set_logging, increment_path, colorstr
-from utils.loss import compute_loss
 from utils.metrics import ap_per_class, ConfusionMatrix
 from utils.plots import plot_images, output_to_target, plot_study_txt
 from utils.torch_utils import select_device, time_synchronized
@@ -36,7 +35,8 @@ def test(data,
          save_hybrid=False,  # for hybrid auto-labelling
          save_conf=False,  # save auto-label confidences
          plots=True,
-         log_imgs=0):  # number of logged images
+         log_imgs=0,  # number of logged images
+         compute_loss=None):
 
     # Initialize/load model and set device
     training = model is not None
@@ -111,8 +111,8 @@ def test(data,
             t0 += time_synchronized() - t
 
             # Compute loss
-            if training:
-                loss += compute_loss([x.float() for x in train_out], targets, model)[1][:3]  # box, obj, cls
+            if compute_loss:
+                loss += compute_loss([x.float() for x in train_out], targets)[1][:3]  # box, obj, cls
 
             # Run NMS
             targets[:, 2:] *= torch.Tensor([width, height, width, height]).to(device)  # to pixels

train.py

@@ -29,7 +29,7 @@ from utils.general import labels_to_class_weights, increment_path, labels_to_ima
     fitness, strip_optimizer, get_latest_run, check_dataset, check_file, check_git_status, check_img_size, \
     check_requirements, print_mutation, set_logging, one_cycle, colorstr
 from utils.google_utils import attempt_download
-from utils.loss import compute_loss
+from utils.loss import ComputeLoss
 from utils.plots import plot_images, plot_labels, plot_results, plot_evolution
 from utils.torch_utils import ModelEMA, select_device, intersect_dicts, torch_distributed_zero_first
 
@@ -227,6 +227,7 @@ def train(hyp, opt, device, tb_writer=None, wandb=None):
     results = (0, 0, 0, 0, 0, 0, 0)  # P, R, [email protected], [email protected], val_loss(box, obj, cls)
     scheduler.last_epoch = start_epoch - 1  # do not move
     scaler = amp.GradScaler(enabled=cuda)
+    compute_loss = ComputeLoss(model)  # init loss class
     logger.info(f'Image sizes {imgsz} train, {imgsz_test} test\n'
                 f'Using {dataloader.num_workers} dataloader workers\n'
                 f'Logging results to {save_dir}\n'
@@ -286,7 +287,7 @@ def train(hyp, opt, device, tb_writer=None, wandb=None):
             # Forward
             with amp.autocast(enabled=cuda):
                 pred = model(imgs)  # forward
-                loss, loss_items = compute_loss(pred, targets.to(device), model)  # loss scaled by batch_size
+                loss, loss_items = compute_loss(pred, targets.to(device))  # loss scaled by batch_size
                 if rank != -1:
                     loss *= opt.world_size  # gradient averaged between devices in DDP mode
                 if opt.quad:
@@ -344,7 +345,8 @@ def train(hyp, opt, device, tb_writer=None, wandb=None):
                                                  dataloader=testloader,
                                                  save_dir=save_dir,
                                                  plots=plots and final_epoch,
-                                                 log_imgs=opt.log_imgs if wandb else 0)
+                                                 log_imgs=opt.log_imgs if wandb else 0,
+                                                 compute_loss=compute_loss)
 
             # Write
             with open(results_file, 'a') as f:

utils/loss.py

@@ -85,119 +85,133 @@ class QFocalLoss(nn.Module):
             return loss
 
 
-def compute_loss(p, targets, model):  # predictions, targets, model
-    device = targets.device
-    lcls, lbox, lobj = torch.zeros(1, device=device), torch.zeros(1, device=device), torch.zeros(1, device=device)
-    tcls, tbox, indices, anchors = build_targets(p, targets, model)  # targets
-    h = model.hyp  # hyperparameters
-
-    # Define criteria
-    BCEcls = nn.BCEWithLogitsLoss(pos_weight=torch.tensor([h['cls_pw']], device=device))  # weight=model.class_weights)
-    BCEobj = nn.BCEWithLogitsLoss(pos_weight=torch.tensor([h['obj_pw']], device=device))
-
-    # Class label smoothing https://arxiv.org/pdf/1902.04103.pdf eqn 3
-    cp, cn = smooth_BCE(eps=0.0)
-
-    # Focal loss
-    g = h['fl_gamma']  # focal loss gamma
-    if g > 0:
-        BCEcls, BCEobj = FocalLoss(BCEcls, g), FocalLoss(BCEobj, g)
-
-    # Losses
-    nt = 0  # number of targets
-    balance = [4.0, 1.0, 0.3, 0.1, 0.03]  # P3-P7
-    for i, pi in enumerate(p):  # layer index, layer predictions
-        b, a, gj, gi = indices[i]  # image, anchor, gridy, gridx
-        tobj = torch.zeros_like(pi[..., 0], device=device)  # target obj
-
-        n = b.shape[0]  # number of targets
-        if n:
-            nt += n  # cumulative targets
-            ps = pi[b, a, gj, gi]  # prediction subset corresponding to targets
-
-            # Regression
-            pxy = ps[:, :2].sigmoid() * 2. - 0.5
-            pwh = (ps[:, 2:4].sigmoid() * 2) ** 2 * anchors[i]
-            pbox = torch.cat((pxy, pwh), 1)  # predicted box
-            iou = bbox_iou(pbox.T, tbox[i], x1y1x2y2=False, CIoU=True)  # iou(prediction, target)
-            lbox += (1.0 - iou).mean()  # iou loss
-
-            # Objectness
-            tobj[b, a, gj, gi] = (1.0 - model.gr) + model.gr * iou.detach().clamp(0).type(tobj.dtype)  # iou ratio
-
-            # Classification
-            if model.nc > 1:  # cls loss (only if multiple classes)
-                t = torch.full_like(ps[:, 5:], cn, device=device)  # targets
-                t[range(n), tcls[i]] = cp
-                lcls += BCEcls(ps[:, 5:], t)  # BCE
-
-            # Append targets to text file
-            # with open('targets.txt', 'a') as file:
-            #     [file.write('%11.5g ' * 4 % tuple(x) + '\n') for x in torch.cat((txy[i], twh[i]), 1)]
-
-        lobj += BCEobj(pi[..., 4], tobj) * balance[i]  # obj loss
-
-    lbox *= h['box']
-    lobj *= h['obj']
-    lcls *= h['cls']
-    bs = tobj.shape[0]  # batch size
-
-    loss = lbox + lobj + lcls
-    return loss * bs, torch.cat((lbox, lobj, lcls, loss)).detach()
-
-
-def build_targets(p, targets, model):
-    # Build targets for compute_loss(), input targets(image,class,x,y,w,h)
-    det = model.module.model[-1] if is_parallel(model) else model.model[-1]  # Detect() module
-    na, nt = det.na, targets.shape[0]  # number of anchors, targets
-    tcls, tbox, indices, anch = [], [], [], []
-    gain = torch.ones(7, device=targets.device)  # normalized to gridspace gain
-    ai = torch.arange(na, device=targets.device).float().view(na, 1).repeat(1, nt)  # same as .repeat_interleave(nt)
-    targets = torch.cat((targets.repeat(na, 1, 1), ai[:, :, None]), 2)  # append anchor indices
-
-    g = 0.5  # bias
-    off = torch.tensor([[0, 0],
-                        [1, 0], [0, 1], [-1, 0], [0, -1],  # j,k,l,m
-                        # [1, 1], [1, -1], [-1, 1], [-1, -1],  # jk,jm,lk,lm
-                        ], device=targets.device).float() * g  # offsets
-
-    for i in range(det.nl):
-        anchors = det.anchors[i]
-        gain[2:6] = torch.tensor(p[i].shape)[[3, 2, 3, 2]]  # xyxy gain
-
-        # Match targets to anchors
-        t = targets * gain
-        if nt:
-            # Matches
-            r = t[:, :, 4:6] / anchors[:, None]  # wh ratio
-            j = torch.max(r, 1. / r).max(2)[0] < model.hyp['anchor_t']  # compare
-            # j = wh_iou(anchors, t[:, 4:6]) > model.hyp['iou_t']  # iou(3,n)=wh_iou(anchors(3,2), gwh(n,2))
-            t = t[j]  # filter
-
-            # Offsets
+class ComputeLoss:
+    # Compute losses
+    def __init__(self, model, autobalance=False):
+        super(ComputeLoss, self).__init__()
+        device = next(model.parameters()).device  # get model device
+        h = model.hyp  # hyperparameters
+
+        # Define criteria
+        BCEcls = nn.BCEWithLogitsLoss(pos_weight=torch.tensor([h['cls_pw']], device=device))
+        BCEobj = nn.BCEWithLogitsLoss(pos_weight=torch.tensor([h['obj_pw']], device=device))
+
+        # Class label smoothing https://arxiv.org/pdf/1902.04103.pdf eqn 3
+        self.cp, self.cn = smooth_BCE(eps=0.0)
+
+        # Focal loss
+        g = h['fl_gamma']  # focal loss gamma
+        if g > 0:
+            BCEcls, BCEobj = FocalLoss(BCEcls, g), FocalLoss(BCEobj, g)
+
+        det = model.module.model[-1] if is_parallel(model) else model.model[-1]  # Detect() module
+        self.balance = {3: [3.67, 1.0, 0.43], 4: [3.78, 1.0, 0.39, 0.22], 5: [3.88, 1.0, 0.37, 0.17, 0.10]}[det.nl]
+        # self.balance = [1.0] * det.nl
+        self.ssi = (det.stride == 16).nonzero(as_tuple=False).item()  # stride 16 index
+        self.BCEcls, self.BCEobj, self.gr, self.hyp, self.autobalance = BCEcls, BCEobj, model.gr, h, autobalance
+        for k in 'na', 'nc', 'nl', 'anchors':
+            setattr(self, k, getattr(det, k))
+
+    def __call__(self, p, targets):  # predictions, targets, model
+        device = targets.device
+        lcls, lbox, lobj = torch.zeros(1, device=device), torch.zeros(1, device=device), torch.zeros(1, device=device)
+        tcls, tbox, indices, anchors = self.build_targets(p, targets)  # targets
+
+        # Losses
+        for i, pi in enumerate(p):  # layer index, layer predictions
+            b, a, gj, gi = indices[i]  # image, anchor, gridy, gridx
+            tobj = torch.zeros_like(pi[..., 0], device=device)  # target obj
+
+            n = b.shape[0]  # number of targets
+            if n:
+                ps = pi[b, a, gj, gi]  # prediction subset corresponding to targets
+
+                # Regression
+                pxy = ps[:, :2].sigmoid() * 2. - 0.5
+                pwh = (ps[:, 2:4].sigmoid() * 2) ** 2 * anchors[i]
+                pbox = torch.cat((pxy, pwh), 1)  # predicted box
+                iou = bbox_iou(pbox.T, tbox[i], x1y1x2y2=False, CIoU=True)  # iou(prediction, target)
+                lbox += (1.0 - iou).mean()  # iou loss
+
+                # Objectness
+                tobj[b, a, gj, gi] = (1.0 - self.gr) + self.gr * iou.detach().clamp(0).type(tobj.dtype)  # iou ratio
+
+                # Classification
+                if self.nc > 1:  # cls loss (only if multiple classes)
+                    t = torch.full_like(ps[:, 5:], self.cn, device=device)  # targets
+                    t[range(n), tcls[i]] = self.cp
+                    lcls += self.BCEcls(ps[:, 5:], t)  # BCE
+
+                # Append targets to text file
+                # with open('targets.txt', 'a') as file:
+                #     [file.write('%11.5g ' * 4 % tuple(x) + '\n') for x in torch.cat((txy[i], twh[i]), 1)]
+
+            obji = self.BCEobj(pi[..., 4], tobj)
+            lobj += obji * self.balance[i]  # obj loss
+            if self.autobalance:
+                self.balance[i] = self.balance[i] * 0.9999 + 0.0001 / obji.detach().item()
+
+        if self.autobalance:
+            self.balance = [x / self.balance[self.ssi] for x in self.balance]
+        lbox *= self.hyp['box']
+        lobj *= self.hyp['obj']
+        lcls *= self.hyp['cls']
+        bs = tobj.shape[0]  # batch size
+
+        loss = lbox + lobj + lcls
+        return loss * bs, torch.cat((lbox, lobj, lcls, loss)).detach()
+
+    def build_targets(self, p, targets):
+        # Build targets for compute_loss(), input targets(image,class,x,y,w,h)
+        na, nt = self.na, targets.shape[0]  # number of anchors, targets
+        tcls, tbox, indices, anch = [], [], [], []
+        gain = torch.ones(7, device=targets.device)  # normalized to gridspace gain
+        ai = torch.arange(na, device=targets.device).float().view(na, 1).repeat(1, nt)  # same as .repeat_interleave(nt)
+        targets = torch.cat((targets.repeat(na, 1, 1), ai[:, :, None]), 2)  # append anchor indices
+
+        g = 0.5  # bias
+        off = torch.tensor([[0, 0],
+                            [1, 0], [0, 1], [-1, 0], [0, -1],  # j,k,l,m
+                            # [1, 1], [1, -1], [-1, 1], [-1, -1],  # jk,jm,lk,lm
+                            ], device=targets.device).float() * g  # offsets
+
+        for i in range(self.nl):
+            anchors = self.anchors[i]
+            gain[2:6] = torch.tensor(p[i].shape)[[3, 2, 3, 2]]  # xyxy gain
+
+            # Match targets to anchors
+            t = targets * gain
+            if nt:
+                # Matches
+                r = t[:, :, 4:6] / anchors[:, None]  # wh ratio
+                j = torch.max(r, 1. / r).max(2)[0] < self.hyp['anchor_t']  # compare
+                # j = wh_iou(anchors, t[:, 4:6]) > model.hyp['iou_t']  # iou(3,n)=wh_iou(anchors(3,2), gwh(n,2))
+                t = t[j]  # filter
+
+                # Offsets
+                gxy = t[:, 2:4]  # grid xy
+                gxi = gain[[2, 3]] - gxy  # inverse
+                j, k = ((gxy % 1. < g) & (gxy > 1.)).T
+                l, m = ((gxi % 1. < g) & (gxi > 1.)).T
+                j = torch.stack((torch.ones_like(j), j, k, l, m))
+                t = t.repeat((5, 1, 1))[j]
+                offsets = (torch.zeros_like(gxy)[None] + off[:, None])[j]
+            else:
+                t = targets[0]
+                offsets = 0
+
+            # Define
+            b, c = t[:, :2].long().T  # image, class
             gxy = t[:, 2:4]  # grid xy
-            gxi = gain[[2, 3]] - gxy  # inverse
-            j, k = ((gxy % 1. < g) & (gxy > 1.)).T
-            l, m = ((gxi % 1. < g) & (gxi > 1.)).T
-            j = torch.stack((torch.ones_like(j), j, k, l, m))
-            t = t.repeat((5, 1, 1))[j]
-            offsets = (torch.zeros_like(gxy)[None] + off[:, None])[j]
-        else:
-            t = targets[0]
-            offsets = 0
-
-        # Define
-        b, c = t[:, :2].long().T  # image, class
-        gxy = t[:, 2:4]  # grid xy
-        gwh = t[:, 4:6]  # grid wh
-        gij = (gxy - offsets).long()
-        gi, gj = gij.T  # grid xy indices
-
-        # Append
-        a = t[:, 6].long()  # anchor indices
-        indices.append((b, a, gj.clamp_(0, gain[3] - 1), gi.clamp_(0, gain[2] - 1)))  # image, anchor, grid indices
-        tbox.append(torch.cat((gxy - gij, gwh), 1))  # box
-        anch.append(anchors[a])  # anchors
-        tcls.append(c)  # class
-
-    return tcls, tbox, indices, anch
+            gwh = t[:, 4:6]  # grid wh
+            gij = (gxy - offsets).long()
+            gi, gj = gij.T  # grid xy indices
+
+            # Append
+            a = t[:, 6].long()  # anchor indices
+            indices.append((b, a, gj.clamp_(0, gain[3] - 1), gi.clamp_(0, gain[2] - 1)))  # image, anchor, grid indices
+            tbox.append(torch.cat((gxy - gij, gwh), 1))  # box
+            anch.append(anchors[a])  # anchors
+            tcls.append(c)  # class
+
+        return tcls, tbox, indices, anch