未验证 提交 46659c8b 编写于 作者: Q qingqing01 提交者: GitHub

Rename filename in dygraph/ppdet/modeling and update changelog. (#1935)

* Rename dir name in modeling

* Update changelog
上级 490886ae
...@@ -2,6 +2,20 @@ ...@@ -2,6 +2,20 @@
## 最新版本信息 ## 最新版本信息
### v2.0-beta(12.20/2020)
- 动态图支持:
- 支持Faster-RCNN, Mask-RCNN, FPN, Cascade Faster/Mask RCNN, YOLOv3和SSD模型,试用版本。
- 模型提升:
- 更新PP-YOLO MobileNetv3 large和small模型,精度提升,并新增裁剪和蒸馏后的模型。
- 新功能:
- 支持VisualDL可视化数据预处理图片。
- Bug修复:
- 修复BlazeFace人脸关键点预测bug。
## 历史版本信息
### v0.5.0(11/2020) ### v0.5.0(11/2020)
- 模型丰富度提升: - 模型丰富度提升:
- 发布SOLOv2系列模型,其中SOLOv2-Light-R50-VD-DCN-FPN 模型在单卡V100上达到 38.6 FPS,加速24% ,COCO验证集精度达到38.8%, 提升2.4绝对百分点。 - 发布SOLOv2系列模型,其中SOLOv2-Light-R50-VD-DCN-FPN 模型在单卡V100上达到 38.6 FPS,加速24% ,COCO验证集精度达到38.8%, 提升2.4绝对百分点。
...@@ -20,10 +34,6 @@ ...@@ -20,10 +34,6 @@
- 新增目标检测全流程教程,新增Jetson平台部署教程。 - 新增目标检测全流程教程,新增Jetson平台部署教程。
## 历史版本信息
### v0.4.0(07/2020) ### v0.4.0(07/2020)
- 模型丰富度提升: - 模型丰富度提升:
- 发布PPYOLO模型,COCO数据集精度达到45.2%,单卡V100预测速度达到72.9 FPS,精度和预测速度优于YOLOv4模型。 - 发布PPYOLO模型,COCO数据集精度达到45.2%,单卡V100预测速度达到72.9 FPS,精度和预测速度优于YOLOv4模型。
......
...@@ -3,23 +3,16 @@ ...@@ -3,23 +3,16 @@
--- ---
## 目录 ## 目录
- [简介](#简介)
- [安装PaddlePaddle](#安装PaddlePaddle) - [安装PaddlePaddle](#安装PaddlePaddle)
- [其他依赖安装](#其他依赖安装) - [其他依赖安装](#其他依赖安装)
- [PaddleDetection](#PaddleDetection) - [PaddleDetection](#PaddleDetection)
## 简介
这份文档介绍了如何安装PaddleDetection及其依赖项(包括PaddlePaddle)。
PaddleDetection的相关信息,请参考[README.md](https://github.com/PaddlePaddle/PaddleDetection/blob/master/README.md).
## 安装PaddlePaddle ## 安装PaddlePaddle
**环境需求:** **环境需求:**
- paddlepaddle >= 2.0rc1
- OS 64位操作系统 - OS 64位操作系统
- Python 3(3.5.1+/3.6/3.7),64位版本 - Python 3(3.5.1+/3.6/3.7),64位版本
- pip/pip3(9.0.1+),64位版本操作系统是 - pip/pip3(9.0.1+),64位版本操作系统是
...@@ -50,7 +43,7 @@ PaddleDetection的相关信息,请参考[README.md](https://github.com/PaddleP ...@@ -50,7 +43,7 @@ PaddleDetection的相关信息,请参考[README.md](https://github.com/PaddleP
**安装Python依赖库:** **安装Python依赖库:**
Python依赖库在[requirements.txt](https://github.com/PaddlePaddle/PaddleDetection/blob/master/requirements.txt)中给出,可通过如下命令安装: Python依赖库在[requirements.txt](../../../requirements.txt)中给出,可通过如下命令安装:
``` ```
pip install -r requirements.txt pip install -r requirements.txt
......
from . import ops from . import ops
from . import bbox from . import bbox
from . import mask from . import mask
from . import backbone from . import backbones
from . import neck from . import necks
from . import head from . import heads
from . import loss from . import losses
from . import architecture from . import architectures
from . import post_process from . import post_process
from . import layers from . import layers
from . import utils from . import utils
...@@ -13,11 +13,11 @@ from . import utils ...@@ -13,11 +13,11 @@ from . import utils
from .ops import * from .ops import *
from .bbox import * from .bbox import *
from .mask import * from .mask import *
from .backbone import * from .backbones import *
from .neck import * from .necks import *
from .head import * from .heads import *
from .loss import * from .losses import *
from .architecture import * from .architectures import *
from .post_process import * from .post_process import *
from .layers import * from .layers import *
from .utils import * from .utils import *
...@@ -26,6 +26,8 @@ from paddle.regularizer import L2Decay ...@@ -26,6 +26,8 @@ from paddle.regularizer import L2Decay
from .name_adapter import NameAdapter from .name_adapter import NameAdapter
from numbers import Integral from numbers import Integral
__all__ = ['ResNet', 'Res5Head']
class ConvNormLayer(nn.Layer): class ConvNormLayer(nn.Layer):
def __init__(self, def __init__(self,
...@@ -317,3 +319,20 @@ class ResNet(nn.Layer): ...@@ -317,3 +319,20 @@ class ResNet(nn.Layer):
if idx in self.return_idx: if idx in self.return_idx:
outs.append(x) outs.append(x)
return outs return outs
@register
class Res5Head(nn.Layer):
def __init__(self, feat_in=1024, feat_out=512):
super(Res5Head, self).__init__()
na = NameAdapter(self)
self.res5_conv = []
self.res5 = self.add_sublayer(
'res5_roi_feat',
Blocks(
feat_in, feat_out, count=3, name_adapter=na, stage_num=5))
self.feat_out = feat_out * 4
def forward(self, roi_feat, stage=0):
y = self.res5(roi_feat)
return y
...@@ -22,9 +22,6 @@ from paddle.regularizer import L2Decay ...@@ -22,9 +22,6 @@ from paddle.regularizer import L2Decay
from ppdet.core.workspace import register from ppdet.core.workspace import register
from ppdet.modeling import ops from ppdet.modeling import ops
from ..backbone.name_adapter import NameAdapter
from ..backbone.resnet import Blocks
@register @register
class TwoFCHead(nn.Layer): class TwoFCHead(nn.Layer):
...@@ -80,23 +77,6 @@ class TwoFCHead(nn.Layer): ...@@ -80,23 +77,6 @@ class TwoFCHead(nn.Layer):
return fc7_relu return fc7_relu
@register
class Res5Head(nn.Layer):
def __init__(self, feat_in=1024, feat_out=512):
super(Res5Head, self).__init__()
na = NameAdapter(self)
self.res5_conv = []
self.res5 = self.add_sublayer(
'res5_roi_feat',
Blocks(
feat_in, feat_out, count=3, name_adapter=na, stage_num=5))
self.feat_out = feat_out * 4
def forward(self, roi_feat, stage=0):
y = self.res5(roi_feat)
return y
@register @register
class BBoxFeat(nn.Layer): class BBoxFeat(nn.Layer):
__inject__ = ['roi_extractor', 'head_feat'] __inject__ = ['roi_extractor', 'head_feat']
......
...@@ -4,7 +4,7 @@ import paddle.nn.functional as F ...@@ -4,7 +4,7 @@ import paddle.nn.functional as F
from paddle import ParamAttr from paddle import ParamAttr
from paddle.regularizer import L2Decay from paddle.regularizer import L2Decay
from ppdet.core.workspace import register from ppdet.core.workspace import register
from ..backbone.darknet import ConvBNLayer from ..backbones.darknet import ConvBNLayer
@register @register
......
...@@ -17,7 +17,7 @@ import paddle.nn as nn ...@@ -17,7 +17,7 @@ import paddle.nn as nn
import paddle.nn.functional as F import paddle.nn.functional as F
from paddle import ParamAttr from paddle import ParamAttr
from ppdet.core.workspace import register, serializable from ppdet.core.workspace import register, serializable
from ..backbone.darknet import ConvBNLayer from ..backbones.darknet import ConvBNLayer
class YoloDetBlock(nn.Layer): class YoloDetBlock(nn.Layer):
......
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