未验证 提交 eeeb92a8 编写于 作者: F Feng Ni 提交者: GitHub

fix denseteacher fcos docs (#7531)

上级 e274ee57
......@@ -21,7 +21,7 @@
variant: d
norm_type: bn
freeze_at: 0
return_idx: [0,1,2,3]
return_idx: [1, 2, 3]
num_stages: 4
lr_mult_list: [0.05, 0.05, 0.1, 0.15]
```
......
......@@ -227,7 +227,7 @@ use_warmup: &use_warmup True
variant: d
norm_type: bn
freeze_at: 0
return_idx: [0,1,2,3]
return_idx: [1, 2, 3]
num_stages: 4
lr_mult_list: [0.05, 0.05, 0.1, 0.15]
```
......
......@@ -28,7 +28,7 @@
variant: d
norm_type: bn
freeze_at: 0
return_idx: [0,1,2,3]
return_idx: [1, 2, 3]
num_stages: 4
lr_mult_list: [0.05, 0.05, 0.1, 0.15]
```
......
......@@ -56,6 +56,7 @@ SemiTrainReader:
object_sizes_boundary: [64, 128, 256, 512]
center_sampling_radius: 1.5
downsample_ratios: [8, 16, 32, 64, 128]
num_shift: 0.5
norm_reg_targets: True
unsup_batch_transforms:
- Permute: {}
......@@ -121,6 +122,7 @@ FCOSHead:
prior_prob: 0.01
norm_reg_targets: True
centerness_on_reg: True
num_shift: 0.5
fcos_loss:
name: FCOSLoss
loss_alpha: 0.25
......@@ -132,7 +134,7 @@ FCOSHead:
name: MultiClassNMS
nms_top_k: 1000
keep_top_k: 100
score_threshold: 0.05
score_threshold: 0.025
nms_threshold: 0.6
......
......@@ -132,7 +132,7 @@ FCOSHead:
name: MultiClassNMS
nms_top_k: 1000
keep_top_k: 100
score_threshold: 0.05
score_threshold: 0.025
nms_threshold: 0.6
......
......@@ -137,7 +137,7 @@ FCOSHead:
name: MultiClassNMS
nms_top_k: 1000
keep_top_k: 100
score_threshold: 0.05
score_threshold: 0.025
nms_threshold: 0.6
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册