未验证 提交 a30e9869 编写于 作者: Z zhiboniu 提交者: GitHub

add dark model_zoo; (#3369)

add config dark_hrnet_w32_384x288.yml;
test=document_fix;
上级 99039025
......@@ -21,8 +21,9 @@
| HigherHRNet+SWAHR | 512 | 32 | 68.9 | [higherhrnet_hrnet_w32_512_swahr.pdparams](https://paddledet.bj.bcebos.com/models/keypoint/higherhrnet_hrnet_w32_512_swahr.pdparams) | [config](./higherhrnet/higherhrnet_hrnet_w32_512_swahr.yml) |
| HRNet | 256x192 | 32 | 76.9 | [hrnet_w32_256x192.pdparams](https://paddledet.bj.bcebos.com/models/keypoint/hrnet_w32_256x192.pdparams) | [config](./hrnet/hrnet_w32_256x192.yml) |
| HRNet | 384x288 | 32 | 77.8 | [hrnet_w32_384x288.pdparams](https://paddledet.bj.bcebos.com/models/keypoint/hrnet_w32_384x288.pdparams) | [config](./hrnet/hrnet_w32_384x288.yml) |
| HRNet+DarkPose | 256x192 | 32 | 78.0 | [dark_hrnet_w32_256x192.pdparams](https://paddledet.bj.bcebos.com/models/keypoint/dark_hrnet_w32_256x192.pdparams) | [config](./hrnet/dark_hrnet_w32_256x192.yml) |
| HRNet+DarkPose | 384x288 | 32 | 78.3 | [dark_hrnet_w32_384x288.pdparams](https://paddledet.bj.bcebos.com/models/keypoint/dark_hrnet_w32_384x288.pdparams) | [config](./hrnet/dark_hrnet_w32_384x288.yml) |
备注: Top-Down模型测试AP结果基于GroundTruth标注框
## 快速开始
......
use_gpu: true
log_iter: 5
save_dir: output
snapshot_epoch: 10
weights: output/hrnet_w32_384x288/model_final
epoch: 210
num_joints: &num_joints 17
pixel_std: &pixel_std 200
metric: KeyPointTopDownCOCOEval
num_classes: 1
train_height: &train_height 384
train_width: &train_width 288
trainsize: &trainsize [*train_width, *train_height]
hmsize: &hmsize [72, 96]
flip_perm: &flip_perm [[1, 2], [3, 4], [5, 6], [7, 8], [9, 10], [11, 12], [13, 14], [15, 16]]
#####model
architecture: TopDownHRNet
pretrain_weights: https://paddledet.bj.bcebos.com/models/pretrained/Trunc_HRNet_W32_C_pretrained.pdparams
TopDownHRNet:
backbone: HRNet
post_process: HRNetPostProcess
flip_perm: *flip_perm
num_joints: *num_joints
width: &width 32
loss: KeyPointMSELoss
flip: true
HRNet:
width: *width
freeze_at: -1
freeze_norm: false
return_idx: [0]
KeyPointMSELoss:
use_target_weight: true
#####optimizer
LearningRate:
base_lr: 0.0005
schedulers:
- !PiecewiseDecay
milestones: [170, 200]
gamma: 0.1
- !LinearWarmup
start_factor: 0.001
steps: 1000
OptimizerBuilder:
optimizer:
type: Adam
regularizer:
factor: 0.0
type: L2
#####data
TrainDataset:
!KeypointTopDownCocoDataset
image_dir: train2017
anno_path: annotations/person_keypoints_train2017.json
dataset_dir: dataset/coco
num_joints: *num_joints
trainsize: *trainsize
pixel_std: *pixel_std
use_gt_bbox: True
EvalDataset:
!KeypointTopDownCocoDataset
image_dir: val2017
anno_path: annotations/person_keypoints_val2017.json
dataset_dir: dataset/coco
bbox_file: bbox.json
num_joints: *num_joints
trainsize: *trainsize
pixel_std: *pixel_std
use_gt_bbox: True
image_thre: 0.0
TestDataset:
!ImageFolder
anno_path: dataset/coco/keypoint_imagelist.txt
worker_num: 2
global_mean: &global_mean [0.485, 0.456, 0.406]
global_std: &global_std [0.229, 0.224, 0.225]
TrainReader:
sample_transforms:
- RandomFlipHalfBodyTransform:
scale: 0.5
rot: 40
num_joints_half_body: 8
prob_half_body: 0.3
pixel_std: *pixel_std
trainsize: *trainsize
upper_body_ids: [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10]
flip_pairs: *flip_perm
- TopDownAffine:
trainsize: *trainsize
- ToHeatmapsTopDown_DARK:
hmsize: *hmsize
sigma: 2
batch_transforms:
- NormalizeImage:
mean: *global_mean
std: *global_std
is_scale: true
- Permute: {}
batch_size: 64
shuffle: true
drop_last: false
EvalReader:
sample_transforms:
- TopDownAffine:
trainsize: *trainsize
- ToHeatmapsTopDown_DARK:
hmsize: *hmsize
sigma: 2
batch_transforms:
- NormalizeImage:
mean: *global_mean
std: *global_std
is_scale: true
- Permute: {}
batch_size: 16
drop_empty: false
TestReader:
sample_transforms:
- Decode: {}
- TopDownEvalAffine:
trainsize: *trainsize
- NormalizeImage:
mean: *global_mean
std: *global_std
is_scale: true
- Permute: {}
batch_size: 1
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