From b7dd8133b382a5e4500bf3a077acf8215e3fd536 Mon Sep 17 00:00:00 2001 From: shangliang Xu Date: Thu, 31 Mar 2022 13:33:20 +0800 Subject: [PATCH] [PPYOLOE] update doc, params, flops (#5543) --- configs/ppyoloe/README.md | 14 +++++++------- configs/ppyoloe/README_cn.md | 14 +++++++------- ppdet/engine/trainer.py | 3 ++- ppdet/modeling/backbones/cspresnet.py | 8 +++++--- 4 files changed, 21 insertions(+), 18 deletions(-) diff --git a/configs/ppyoloe/README.md b/configs/ppyoloe/README.md index 365c18e76..1fca3102e 100644 --- a/configs/ppyoloe/README.md +++ b/configs/ppyoloe/README.md @@ -9,7 +9,7 @@ English | [简体中文](README_cn.md) - [Appendix](#Appendix) ## Introduction -PP-YOLOE is an excellent single-stage anchor-free model based on PP-YOLOv2, surpassing a variety of popular yolo models. PP-YOLOE has a series of models, named s/m/l/x, which are configured through width multiplier and depth multiplier. PP-YOLOE avoids using special operators, such as deformable convolution or matrix nms, to be deployed friendly on various hardware. For more details, please refer to our report. +PP-YOLOE is an excellent single-stage anchor-free model based on PP-YOLOv2, surpassing a variety of popular yolo models. PP-YOLOE has a series of models, named s/m/l/x, which are configured through width multiplier and depth multiplier. PP-YOLOE avoids using special operators, such as deformable convolution or matrix nms, to be deployed friendly on various hardware. For more details, please refer to our [report](https://arxiv.org/abs/2203.16250).
@@ -24,12 +24,12 @@ PP-YOLOE is composed of following methods: - [SiLU activation function](https://arxiv.org/abs/1710.05941) ## Model Zoo -| Model | GPU number | images/GPU | backbone | input shape | Box APval | Box APtest | V100 FP32(FPS) | V100 TensorRT FP16(FPS) | download | config | -|:------------------------:|:-------:|:-------------:|:----------:| :-------:| :------------------: | :-------------------: | :------------: | :---------------------: | :------: | :------: | -| PP-YOLOE-s | 8 | 32 | cspresnet-s | 640 | 42.7 | 43.1 | 208.3 | 333.3 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_s_300e_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyoloe/ppyoloe_crn_s_300e_coco.yml) | -| PP-YOLOE-m | 8 | 32 | cspresnet-m | 640 | 48.6 | 48.9 | 123.4 | 208.3 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_m_300e_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyoloe/ppyoloe_crn_m_300e_coco.yml) | -| PP-YOLOE-l | 8 | 24 | cspresnet-l | 640 | 50.9 | 51.4 | 78.1 | 149.2 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_l_300e_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyoloe/ppyoloe_crn_l_300e_coco.yml) | -| PP-YOLOE-x | 8 | 16 | cspresnet-x | 640 | 51.9 | 52.2 | 45.0 | 95.2 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_x_300e_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyoloe/ppyoloe_crn_x_300e_coco.yml) | +| Model | GPU number | images/GPU | backbone | input shape | Box APval | Box APtest | Params(M) | FLOPs(G) | V100 FP32(FPS) | V100 TensorRT FP16(FPS) | download | config | +|:------------------------:|:-------:|:----------:|:----------:| :-------:| :------------------: | :-------------------: |:---------:|:--------:| :------------: | :---------------------: | :------: | :------: | +| PP-YOLOE-s | 8 | 32 | cspresnet-s | 640 | 42.7 | 43.1 | 7.93 | 17.36 | 208.3 | 333.3 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_s_300e_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyoloe/ppyoloe_crn_s_300e_coco.yml) | +| PP-YOLOE-m | 8 | 28 | cspresnet-m | 640 | 48.6 | 48.9 | 23.43 | 49.91 | 123.4 | 208.3 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_m_300e_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyoloe/ppyoloe_crn_m_300e_coco.yml) | +| PP-YOLOE-l | 8 | 20 | cspresnet-l | 640 | 50.9 | 51.4 | 52.20 | 110.07 | 78.1 | 149.2 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_l_300e_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyoloe/ppyoloe_crn_l_300e_coco.yml) | +| PP-YOLOE-x | 8 | 16 | cspresnet-x | 640 | 51.9 | 52.2 | 98.42 | 206.59 | 45.0 | 95.2 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_x_300e_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyoloe/ppyoloe_crn_x_300e_coco.yml) | **Notes:** diff --git a/configs/ppyoloe/README_cn.md b/configs/ppyoloe/README_cn.md index 51159b483..72050b6cb 100644 --- a/configs/ppyoloe/README_cn.md +++ b/configs/ppyoloe/README_cn.md @@ -9,7 +9,7 @@ - [附录](#附录) ## 简介 -PP-YOLOE是基于PP-YOLOv2的卓越的单阶段Anchor-free模型，超越了多种流行的yolo模型。PP-YOLOE有一系列的模型，即s/m/l/x，可以通过width multiplier和depth multiplier配置。PP-YOLOE避免使用诸如deformable convolution或者matrix nms之类的特殊算子，以使其能轻松地部署在多种多样的硬件上。更多细节可以参考我们的report。 +PP-YOLOE是基于PP-YOLOv2的卓越的单阶段Anchor-free模型，超越了多种流行的yolo模型。PP-YOLOE有一系列的模型，即s/m/l/x，可以通过width multiplier和depth multiplier配置。PP-YOLOE避免使用诸如deformable convolution或者matrix nms之类的特殊算子，以使其能轻松地部署在多种多样的硬件上。更多细节可以参考我们的[report](https://arxiv.org/abs/2203.16250)。
@@ -24,12 +24,12 @@ PP-YOLOE由以下方法组成 - [SiLU激活函数](https://arxiv.org/abs/1710.05941) ## 模型库 -| 模型 | GPU个数 | 每GPU图片个数 | 骨干网络 | 输入尺寸 | Box APval | Box APtest | V100 FP32(FPS) | V100 TensorRT FP16(FPS) | 模型下载 | 配置文件 | -|:------------------------:|:-------:|:-------------:|:----------:| :-------:| :------------------: | :-------------------: | :------------: | :---------------------: | :------: | :------: | -| PP-YOLOE-s | 8 | 32 | cspresnet-s | 640 | 42.7 | 43.1 | 208.3 | 333.3 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_s_300e_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyoloe/ppyoloe_crn_s_300e_coco.yml) | -| PP-YOLOE-m | 8 | 32 | cspresnet-m | 640 | 48.6 | 48.9 | 123.4 | 208.3 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_m_300e_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyoloe/ppyoloe_crn_m_300e_coco.yml) | -| PP-YOLOE-l | 8 | 24 | cspresnet-l | 640 | 50.9 | 51.4 | 78.1 | 149.2 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_l_300e_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyoloe/ppyoloe_crn_l_300e_coco.yml) | -| PP-YOLOE-x | 8 | 16 | cspresnet-x | 640 | 51.9 | 52.2 | 45.0 | 95.2 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_x_300e_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyoloe/ppyoloe_crn_x_300e_coco.yml) | +| 模型 | GPU个数 | 每GPU图片个数 | 骨干网络 | 输入尺寸 | Box APval | Box APtest | Params(M) | FLOPs(G) | V100 FP32(FPS) | V100 TensorRT FP16(FPS) | 模型下载 | 配置文件 | +|:------------------------:|:-------:|:--------:|:----------:| :-------:| :------------------: | :-------------------: |:---------:|:--------:|:---------------:| :---------------------: | :------: | :------: | +| PP-YOLOE-s | 8 | 32 | cspresnet-s | 640 | 42.7 | 43.1 | 7.93 | 17.36 | 208.3 | 333.3 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_s_300e_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyoloe/ppyoloe_crn_s_300e_coco.yml) | +| PP-YOLOE-m | 8 | 28 | cspresnet-m | 640 | 48.6 | 48.9 | 23.43 | 49.91 | 123.4 | 208.3 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_m_300e_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyoloe/ppyoloe_crn_m_300e_coco.yml) | +| PP-YOLOE-l | 8 | 20 | cspresnet-l | 640 | 50.9 | 51.4 | 52.20 | 110.07 | 78.1 | 149.2 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_l_300e_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyoloe/ppyoloe_crn_l_300e_coco.yml) | +| PP-YOLOE-x | 8 | 16 | cspresnet-x | 640 | 51.9 | 52.2 | 98.42 | 206.59 | 45.0 | 95.2 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_x_300e_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyoloe/ppyoloe_crn_x_300e_coco.yml) | **注意:** diff --git a/ppdet/engine/trainer.py b/ppdet/engine/trainer.py index fa9167f05..18c87c2bc 100644 --- a/ppdet/engine/trainer.py +++ b/ppdet/engine/trainer.py @@ -373,7 +373,8 @@ class Trainer(object): # enabel auto mixed precision mode if self.cfg.get('amp', False): scaler = amp.GradScaler( - enable=self.cfg.use_gpu or self.cfg.use_npu, init_loss_scaling=1024) + enable=self.cfg.use_gpu or self.cfg.use_npu, + init_loss_scaling=1024) self.status.update({ 'epoch_id': self.start_epoch, diff --git a/ppdet/modeling/backbones/cspresnet.py b/ppdet/modeling/backbones/cspresnet.py index 4e0916320..ce8450a86 100644 --- a/ppdet/modeling/backbones/cspresnet.py +++ b/ppdet/modeling/backbones/cspresnet.py @@ -96,6 +96,8 @@ class RepVggBlock(nn.Layer): kernel, bias = self.get_equivalent_kernel_bias() self.conv.weight.set_value(kernel) self.conv.bias.set_value(bias) + self.__delattr__('conv1') + self.__delattr__('conv2') def get_equivalent_kernel_bias(self): kernel3x3, bias3x3 = self._fuse_bn_tensor(self.conv1) @@ -176,7 +178,7 @@ class CSPResStage(nn.Layer): self.conv_down = None self.conv1 = ConvBNLayer(ch_mid, ch_mid // 2, 1, act=act) self.conv2 = ConvBNLayer(ch_mid, ch_mid // 2, 1, act=act) - self.blocks = nn.Sequential(* [ + self.blocks = nn.Sequential(*[ block_fn( ch_mid // 2, ch_mid // 2, act=act, shortcut=True) for i in range(n) @@ -252,9 +254,9 @@ class CSPResNet(nn.Layer): act=act))) n = len(channels) - 1 - self.stages = nn.Sequential(* [(str(i), CSPResStage( + self.stages = nn.Sequential(*[(str(i), CSPResStage( BasicBlock, channels[i], channels[i + 1], layers[i], 2, act=act)) - for i in range(n)]) + for i in range(n)]) self._out_channels = channels[1:] self._out_strides = [4, 8, 16, 32] -- GitLab