Skip to content

Commit

Permalink
[Refactor] Refactor Deploy Code (PaddlePaddle#2881)
Browse files Browse the repository at this point in the history 
* Refactor deploy code
* Update RTFormer and FCN_UHRNetW18_Small TIPC configs
* Add 'add_rule' interface
* Polish README
* Refactor infer API
  • Loading branch information
@Bobholamovic
Bobholamovic authored Jan 3, 2023
1 parent fb37c0c commit 8645221
Show file tree
Hide file tree
Showing 20 changed files with 247 additions and 270 deletions.
40 changes: 20 additions & 20 deletions README_EN.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -6,7 +6,7 @@ English | [简体中文](README_CN.md)
<img src="./docs/images/paddleseg_logo.png" align="middle" width = "500" />
</p>

**A High-Efficient Development Toolkit for Image Segmentation based on [PaddlePaddle](https://github.com/paddlepaddle/paddle).**
**A High-Efficient Development Toolkit for Image Segmentation Based on [PaddlePaddle](https://github.com/paddlepaddle/paddle).**

[![License](https://img.shields.io/badge/license-Apache%202-blue.svg)](LICENSE)
[![Version](https://img.shields.io/github/release/PaddlePaddle/PaddleSeg.svg)](https://github.com/PaddlePaddle/PaddleSeg/releases)
Expand All @@ -23,22 +23,22 @@ English | [简体中文](README_CN.md)

## <img src="./docs/images/seg_news_icon.png" width="20"/> News
<ul class="nobull">
<li>[2022-11-30] :fire: PaddleSeg v2.7 is released! More details in <a href="https://github.com/PaddlePaddle/PaddleSeg/releases">Release Notes</a>.</li>
<li>[2022-11-30] :fire: PaddleSeg v2.7 is released! Check more details in <a href="https://github.com/PaddlePaddle/PaddleSeg/releases">Release Notes</a>.</li>
<ul>
<li>Release <a href="./Matting/">PP-MattingV2</a>, a real-time human matting model with SOTA performance. Compared to previous models, the mean error is reduced by 17.91%, the inference speed is improved by 44.6% on GPU. </li>
<li>Release <a href="./contrib/MedicalSeg/">MedicalSegV2</a>, a superior 3D medical image segmentation solution, including an intelligent annotation toolkit called EISeg-Med3D, several state-of-the-art models and an optimized nnUNet-D with high performance.</li>
<li>Release <a href="./configs/rtformer/">RTFormer</a>, a real-time semantic segmentation model accepted by NeurIPS 2022. RTFormer combines the advantages of CNN and Transformer modules, and it achieves SOTA trade-off between performance and efficiency on several datasets.</li>
<li>Release <a href="./contrib/MedicalSeg/">MedicalSegV2</a>, a superior 3D medical image segmentation solution, including an intelligent annotation toolkit called EISeg-Med3D, several state-of-the-art models, and an optimized nnUNet-D with high performance.</li>
<li>Release <a href="./configs/rtformer/">RTFormer</a>, a real-time semantic segmentation model (the paper has been accepted by NeurIPS 2022). RTFormer combines the advantages of CNN and Transformer modules, and it achieves SOTA trade-off between performance and efficiency on several datasets.</li>
</ul>
<li>[2022-07-20] PaddleSeg v2.6 released a real-time human segmentation SOTA solution <a href="./contrib/PP-HumanSeg">PP-HumanSegV2</a>, a stable-version semi-automatic segmentation annotation <a href="./EISeg">EISeg v1.0</a>, a pseudo label pre-training method PSSL and the source code of PP-MattingV1. </li>
<li>[2022-07-20] PaddleSeg v2.6 released a real-time human segmentation SOTA solution <a href="./contrib/PP-HumanSeg">PP-HumanSegV2</a>, a stable-version semi-automatic segmentation annotation tool <a href="./EISeg">EISeg v1.0</a>, a pseudo label pre-training method PSSL, and the source code of PP-MattingV1. </li>
<li>[2022-04-20] PaddleSeg v2.5 released a real-time semantic segmentation model <a href="./configs/pp_liteseg">PP-LiteSeg</a>, a trimap-free image matting model PP-MattingV1, and an easy-to-use solution for 3D medical image segmentation MedicalSegV1.</li>
<li>[2022-01-20] We release PaddleSeg v2.4 with EISeg v0.4, and PP-HumanSegV1 including open-sourced dataset <a href="./contrib/PP-HumanSeg/paper.md#pp-humanseg14k-a-large-scale-teleconferencing-video-dataset">PP-HumanSeg14K</a>. </li>
<li>[2022-01-20] We release PaddleSeg v2.4 with EISeg v0.4, and PP-HumanSegV1 including an open-sourced dataset <a href="./contrib/PP-HumanSeg/paper.md#pp-humanseg14k-a-large-scale-teleconferencing-video-dataset">PP-HumanSeg14K</a>. </li>

</ul>


## <img src="https://user-images.githubusercontent.com/48054808/157795569-9fc77c85-732f-4870-9be0-99a7fe2cff27.png" width="20"/> Introduction

PaddleSeg is an end-to-end high-efficent development toolkit for image segmentation based on PaddlePaddle, which helps both developers and researchers in the whole process of designing segmentation models, training models, optimizing performance and inference speed, and deploying models. A lot of well-trained models and various real-world applications in both industry and academia help users conveniently build hands-on experiences in image segmentation.
PaddleSeg is an end-to-end high-efficent development toolkit for image segmentation based on PaddlePaddle, which helps both developers and researchers in the whole process of designing segmentation models, training models, optimizing performance and inference speed, and deploying models. A lot of well-trained models and various real-world applications in both industry and academia help users conveniently build hands-on experiences in image segmentation.

<div align="center">
<img src="https://github.com/shiyutang/files/raw/main/teasor_new.gif" width = "800" />
Expand All @@ -47,22 +47,22 @@ PaddleSeg is an end-to-end high-efficent development toolkit for image segmentat

## <img src="./docs/images/feature.png" width="20"/> Features

* **High-Performance Model**: Following the state of the art segmentation methods and use the high-performance backbone, we provide 40+ models and 140+ high-quality pre-training models, which are better than other open-source implementations.
* **High-Performance Model**: Following the state of the art segmentation methods and using high-performance backbone networks, we provide 40+ models and 140+ high-quality pre-training models, which are better than other open-source implementations.

* **High Efficiency**: PaddleSeg provides multi-process asynchronous I/O, multi-card parallel training, evaluation, and other acceleration strategies, combined with the memory optimization function of the PaddlePaddle, which can greatly reduce the training overhead of the segmentation model, all this allowing developers to lower cost and more efficiently train image segmentation model.
* **High Efficiency**: PaddleSeg provides multi-process asynchronous I/O, multi-card parallel training, evaluation, and other acceleration strategies, combined with the memory optimization function of the PaddlePaddle, which can greatly reduce the training overhead of the segmentation model, all these allowing developers to train image segmentation models more efficiently and at a lower cost.

* **Modular Design**: We desigin PaddleSeg with the modular design philosophy. Therefore, based on actual application scenarios, developers can assemble diversified training configurations with *data enhancement strategies*, *segmentation models*, *backbone networks*, *loss functions* and other different components to meet different performance and accuracy requirements.
* **Modular Design**: We build PaddleSeg with the modular design philosophy. Therefore, based on actual application scenarios, developers can assemble diversified training configurations with *data augmentation strategies*, *segmentation models*, *backbone networks*, *loss functions*, and other different components to meet different performance and accuracy requirements.

* **Complete Flow**: PaddleSeg support image labeling, model designing, model training, model compression and model deployment. With the help of PaddleSeg, developers can easily finish all taskes.
* **Complete Flow**: PaddleSeg supports image labeling, model designing, model training, model compression, and model deployment. With the help of PaddleSeg, developers can easily finish all tasks in the entire workflow.

<div align="center">
<img src="https://user-images.githubusercontent.com/14087480/176402154-390e5815-1a87-41be-9374-9139c632eb66.png" width = "800" />
</div>

## <img src="./docs/images/chat.png" width="20"/> Community

* If you have any questions, suggestions and feature requests, please create an issues in [GitHub Issues](https://github.com/PaddlePaddle/PaddleSeg/issues).
* Welcome to scan the following QR code and join paddleseg wechat group to communicate with us.
* If you have any questions, suggestions or feature requests, please do not hesitate to create an issue in [GitHub Issues](https://github.com/PaddlePaddle/PaddleSeg/issues).
* Please scan the following QR code to join PaddleSeg WeChat group to communicate with us:
<div align="center">
<img src="https://user-images.githubusercontent.com/48433081/174770518-e6b5319b-336f-45d9-9817-da12b1961fb1.jpg" width = "200" />
</div>
Expand Down Expand Up @@ -321,7 +321,7 @@ PaddleSeg is an end-to-end high-efficent development toolkit for image segmentat
| CCNet | ResNet101_OS8 | 80.95 | 3.24 | [yml](./configs/ccnet/) |

Note that:
* Test the inference speed on Nvidia GPU V100: use PaddleInference Python API, enable TensorRT, the data type is FP32, the dimension of input is 1x3x1024x2048.
* We test the inference speed on Nvidia GPU V100. We use PaddleInference Python API with TensorRT enabled. The data type is FP32, and the shape of input tensor is 1x3x1024x2048.

</details>

Expand All @@ -344,8 +344,8 @@ Note that:
| SFNet | ResNet18_OS8 | 78.72 | *10.72* | - | [yml](./configs/sfnet/) |

Note that:
* Test the inference speed on Nvidia GPU V100: use PaddleInference Python API, enable TensorRT, the data type is FP32, the dimension of input is 1x3x1024x2048.
* Test the inference speed on Snapdragon 855: use PaddleLite CPP API, 1 thread, the dimension of input is 1x3x256x256.
* We test the inference speed on Nvidia GPU V100. We use PaddleInference Python API with TensorRT enabled. The data type is FP32, and the shape of input tensor is 1x3x1024x2048.
* We test the inference speed on Snapdragon 855. We use PaddleLite CPP API with 1 thread, and the shape of input tensor is 1x3x256x256.

</details>

Expand All @@ -364,8 +364,8 @@ Note that:
| MobileSeg | GhostNet_x1_0 | 71.88 | *35.58* | 38.74 | [yml](./configs/mobileseg/) |

Note that:
* Test the inference speed on Nvidia GPU V100: use PaddleInference Python API, enable TensorRT, the data type is FP32, the dimension of input is 1x3x1024x2048.
* Test the inference speed on Snapdragon 855: use PaddleLite CPP API, 1 thread, the dimension of input is 1x3x256x256.
* We test the inference speed on Nvidia GPU V100. We use PaddleInference Python API with TensorRT enabled. The data type is FP32, and the shape of input tensor is 1x3x1024x2048.
* We test the inference speed on Snapdragon 855. We use PaddleLite CPP API with 1 thread, and the shape of input tensor is 1x3x256x256.

</details>

Expand Down Expand Up @@ -394,7 +394,7 @@ Note that:
* [Export Inference Model](./docs/model_export.md)
* [Export ONNX Model](./docs/model_export_onnx.md)

* Model Deploy
* Model Deployment
* [Paddle Inference (Python)](./docs/deployment/inference/python_inference.md)
* [Paddle Inference (C++)](./docs/deployment/inference/cpp_inference.md)
* [Paddle Lite](./docs/deployment/lite/lite.md)
Expand All @@ -409,7 +409,7 @@ Note that:
* Model Compression
* [Quantization](./docs/deployment/slim/quant/quant.md)
* [Distillation](./docs/deployment/slim/distill/distill.md)
* [Prune](./docs/deployment/slim/prune/prune.md)
* [Pruning](./docs/deployment/slim/prune/prune.md)

* [FAQ](./docs/faq/faq/faq.md)

Expand Down
10 changes: 2 additions & 8 deletions deploy/python/collect_dynamic_shape.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -13,20 +13,14 @@
# limitations under the License.

import argparse
import codecs
import os
import sys

import yaml
import numpy as np
from paddle.inference import create_predictor, PrecisionType
from paddle.inference import create_predictor
from paddle.inference import Config as PredictConfig

LOCAL_PATH = os.path.dirname(os.path.abspath(__file__))
sys.path.append(os.path.join(LOCAL_PATH, '..', '..'))

from paddleseg.utils import logger, get_image_list, progbar
from infer import DeployConfig
from paddleseg.deploy.infer import DeployConfig
"""
Load images and run the model, it collects and saves dynamic shapes,
which are used in deployment with TRT.
Expand Down
Loading

0 comments on commit 8645221

Please sign in to comment.