@@ -12428,7 +12428,7 @@ multi_box_head
12428
12428
.. code- block:: python
12429
12429
12430
12430
mbox_locs, mbox_confs, box, var = fluid.layers.multi_box_head(
12431
- inputs = [conv1, conv2, conv3, conv4, conv5, conv5 ],
12431
+ inputs = [conv1, conv2, conv3, conv4, conv5, conv6 ],
12432
12432
image = images,
12433
12433
num_classes = 21 ,
12434
12434
min_ratio = 20 ,
@@ -12914,7 +12914,7 @@ yolo_box
12914
12914
yolov3_loss
12915
12915
------------------------------ -
12916
12916
12917
- .. py:function:: paddle.fluid.layers.yolov3_loss(x, gtbox, gtlabel , anchors, anchor_mask, class_num, ignore_thresh, downsample_ratio, gtscore = None , use_label_smooth = True , name = None )
12917
+ .. py:function:: paddle.fluid.layers.yolov3_loss(x, gt_box, gt_label , anchors, anchor_mask, class_num, ignore_thresh, downsample_ratio, gt_score = None , use_label_smooth = True , name = None )
12918
12918
12919
12919
该运算通过给定的预测结果和真实框生成yolov3损失。
12920
12920
@@ -12965,15 +12965,15 @@ yolov3_loss
12965
12965
12966
12966
参数:
12967
12967
- ** x** (Variable) – YOLOv3损失运算的输入张量,这是一个形状为[N,C,H,W]的四维张量。H和W应该相同,第二维(C)存储框的位置信息,以及每个anchor box的置信度得分和one- hot分类
12968
- - ** gtbox ** (Variable) – 真实框,应该是[N,B,4 ]的形状。第三维用来承载x、y、w、h,x、y、w、h应该是输入图像相对值。 N是batch size,B是图像中所含有的的最多的box数目
12969
- - ** gtlabel ** (Variable) – 真实框的类id ,应该形为[N,B]。
12968
+ - ** gt_box ** (Variable) – 真实框,应该是[N,B,4 ]的形状。第三维用来承载x、y、w、h,x、y、w、h应该是输入图像相对值。 N是batch size,B是图像中所含有的的最多的box数目
12969
+ - ** gt_label ** (Variable) – 真实框的类id ,应该形为[N,B]。
12970
12970
- ** anchors** (list | tuple ) – 指定anchor框的宽度和高度,它们将逐对进行解析
12971
12971
- ** anchor_mask** (list | tuple ) – 当前YOLOv3损失计算中使用的anchor的mask索引
12972
12972
- ** class_num** (int ) – 要预测的类数
12973
12973
- ** ignore_thresh** (float ) – 一定条件下忽略某框置信度损失的忽略阈值
12974
12974
- ** downsample_ratio** (int ) – 从网络输入到YOLOv3 loss输入的下采样率,因此应为第一,第二和第三个YOLOv3损失运算设置32 ,16 ,8
12975
12975
- ** name** (string) – yolov3损失层的命名
12976
- - ** gtscore ** (Variable) - 真实框的混合得分,形为[N,B]。 默认None 。
12976
+ - ** gt_score ** (Variable) - 真实框的混合得分,形为[N,B]。 默认None 。
12977
12977
- ** use_label_smooth** (bool ) - 是否使用平滑标签。 默认为True
12978
12978
12979
12979
@@ -12996,13 +12996,13 @@ yolov3_loss
12996
12996
.. code- block:: python
12997
12997
12998
12998
x = fluid.layers.data(name = ' x' , shape = [255 , 13 , 13 ], dtype = ' float32' )
12999
- gtbox = fluid.layers.data(name = ' gtbox' , shape = [6 , 4 ], dtype = ' float32' )
13000
- gtlabel = fluid.layers.data(name = ' gtlabel' , shape = [6 ], dtype = ' int32' )
13001
- gtscore = fluid.layers.data(name = ' gtscore' , shape = [6 ], dtype = ' float32' )
12999
+ gt_box = fluid.layers.data(name = ' gtbox' , shape = [6 , 4 ], dtype = ' float32' )
13000
+ gt_label = fluid.layers.data(name = ' gtlabel' , shape = [6 ], dtype = ' int32' )
13001
+ gt_score = fluid.layers.data(name = ' gtscore' , shape = [6 ], dtype = ' float32' )
13002
13002
anchors = [10 , 13 , 16 , 30 , 33 , 23 , 30 , 61 , 62 , 45 , 59 , 119 , 116 , 90 , 156 , 198 , 373 , 326 ]
13003
13003
anchor_mask = [0 , 1 , 2 ]
13004
- loss = fluid.layers.yolov3_loss(x = x, gtbox = gtbox, gtlabel = gtlabel ,
13005
- gtscore = gtscore , anchors = anchors,
13004
+ loss = fluid.layers.yolov3_loss(x = x, gt_box = gt_box, gt_label = gt_label ,
13005
+ gt_score = gt_score , anchors = anchors,
13006
13006
anchor_mask = anchor_mask, class_num = 80 ,
13007
13007
ignore_thresh = 0.7 , downsample_ratio = 32 )
13008
13008
0 commit comments