polish

README.md

@@ -89,6 +89,7 @@ python train.py --config configs/quick_start/bisenet_optic_disc_512x512_1k.yml
 * [API Tutorial](https://aistudio.baidu.com/aistudio/projectdetail/1339458)
 * [Data Preparation](./docs/data_prepare.md)
 * [Training Configuration](./configs/)
+* [Loss Usage](./docs/loss_usage.md)
 * [API References](./docs/apis)
 * [Add New Components](./docs/add_new_model.md)
 

README_CN.md

@@ -90,6 +90,7 @@ python train.py --config configs/quick_start/bisenet_optic_disc_512x512_1k.yml
 * [API使用教程](https://aistudio.baidu.com/aistudio/projectdetail/1339458)
 * [数据集准备](./docs/data_prepare.md)
 * [配置项](./configs/)
+* [Loss使用](./docs/loss_usage.md)
 * [API参考](./docs/apis)
 * [添加新组件](./docs/add_new_model.md)
 

docs/loss_usage.md

@@ -0,0 +1,4 @@
+# Loss usage
+
+- [Lovasz loss](lovasz_loss.md)
+- To be continued

paddleseg/models/losses/lovasz_loss.py

@@ -46,14 +46,10 @@ def forward(self, logits, labels):
 
         Args:
             logits: [N, C, H, W] Tensor, logits at each prediction (between -\infty and +\infty)
-        labels: [N, 1, H, W] or [N, H, W] Tensor, ground truth labels (between 0 and C - 1)
+            labels: [N, 1, H, W] or [N, H, W] Tensor, ground truth labels (between 0 and C - 1)
         """
-        probas = F.softmax(
-            logits, axis=1
-        )  # probas grad [ 0.0000000e+00  0.0000000e+00  0.0000000e+00  0.0000000e+00  -5.5730345e-07]
-        vprobas, vlabels = flatten_probas(
-            probas, labels, self.ignore_index
-        )  # vprobas grad和torch不同，第一个类别是0，torch不是，第一个类别的第一个元素是0.05
+        probas = F.softmax(logits, axis=1)
+        vprobas, vlabels = flatten_probas(probas, labels, self.ignore_index)
         loss = lovasz_softmax_flat(vprobas, vlabels, classes=self.classes)
         return loss
 
@@ -100,10 +96,6 @@ def lovasz_grad(gt_sorted):
 
     if p > 1:  # cover 1-pixel case
         jaccard[1:p] = jaccard[1:p] - jaccard[0:-1]
-        # jaccard0 = paddle.slice(jaccard, axis=[0], starts=[0], ends=[1])
-        # jaccard1 = paddle.slice(jaccard, axis=[0], starts=[1], ends=[len_gt])
-        # jaccard2 = paddle.slice(jaccard, axis=[0], starts=[0], ends=[-1])
-        # jaccard = paddle.concat([jaccard0, jaccard1 - jaccard2], axis=0)
     return jaccard
 
 
@@ -148,14 +140,11 @@ def flatten_binary_scores(scores, labels, ignore=None):
     if ignore is None:
         return scores, labels
     valid = labels != ignore
-    # ignore = paddle.cast(ignore, 'int32')
     valid_mask = paddle.reshape(valid, (-1, 1))
     indexs = paddle.nonzero(valid_mask)
     indexs.stop_gradient = True
     vscores = paddle.gather(scores, indexs[:, 0])
     vlabels = paddle.gather(labels, indexs[:, 0])
-    # vscores = paddle.squeeze(vscores, axis=1)
-    # vlabels = paddle.squeeze(vlabels, axis=1)
     return vscores, vlabels
 
 
@@ -184,23 +173,23 @@ def lovasz_softmax_flat(probas, labels, classes='present'):
             class_pred = probas[:, 0]
         else:
             class_pred = probas[:, c]
-        errors = paddle.abs(fg - class_pred)  # errors梯度不同
+        errors = paddle.abs(fg - class_pred)
         errors_sorted, perm = paddle.fluid.core.ops.argsort(
             errors, 'axis', 0, 'descending', True)
-        errors_sorted.stop_gradient = False  # errors_sorted梯度相同
+        errors_sorted.stop_gradient = False
 
         fg_sorted = paddle.gather(fg, perm)
         fg_sorted.stop_gradient = True
 
-        grad = lovasz_grad(fg_sorted)  # grad值相同，无梯度
+        grad = lovasz_grad(fg_sorted)
         grad.stop_gradient = True
         loss = paddle.sum(errors_sorted * grad)
-        losses.append(loss)  # loss梯度相同，值相同
+        losses.append(loss)
 
     if len(classes_to_sum) == 1:
         return losses[0]
 
-    losses_tensor = paddle.stack(losses)  # losses_tensor值相同，梯度也相同
+    losses_tensor = paddle.stack(losses)
     mean_loss = paddle.mean(losses_tensor)
     return mean_loss
 
@@ -218,12 +207,9 @@ def flatten_probas(probas, labels, ignore=None):
     if ignore is None:
         return probas, labels
     valid = labels != ignore
-    # valid = paddle.cast(valid, 'int32')
     valid_mask = paddle.reshape(valid, [-1, 1])
     indexs = paddle.nonzero(valid_mask)
     indexs.stop_gradient = True
     vprobas = paddle.gather(probas, indexs[:, 0])
-    # print(probas.shape, vprobas.shape)  # [1789832, 20] [1700971, 20]
     vlabels = paddle.gather(labels, indexs[:, 0])
-    # vlabels = paddle.squeeze(vlabels, axis=1)
     return vprobas, vlabels

paddleseg/models/losses/mixed_loss.py

@@ -21,6 +21,18 @@
 
 @manager.LOSSES.add_component
 class MixedLoss(nn.Layer):
+    """
+    Weighted computations for multiple Loss.
+    The advantage is that mixed loss training can be achieved without changing the networking code.
+
+    Args:
+        losses (list of nn.Layer): A list consisting of multiple loss classes
+        coef (float|int): Weighting coefficient of multiple loss
+
+    Returns:
+        A callable object of MixedLoss.
+    """
+
     def __init__(self, losses, coef):
         super(MixedLoss, self).__init__()
         if not isinstance(losses, list):
@@ -38,16 +50,9 @@ def __init__(self, losses, coef):
         self.coef = coef
 
     def forward(self, logits, labels):
-        final_output = 0
         loss_list = []
+        final_output = 0
         for i, loss in enumerate(self.losses):
             output = loss(logits, labels)
             final_output += output * self.coef[i]
-            loss_list.append(output * self.coef[i])
-        return final_output, loss_list
-
-        # loss_list = []
-        # for i, loss in enumerate(self.losses):
-        #     output = loss(logits, labels)
-        #     loss_list.append(output * self.coef[i])
-        # return loss_list
+        return final_output