PaddleX v2.0.0 rc0

PaddleX - a full-process development tool for PaddleX, supports developers to quickly implement actual industrial projects in the form of low code.

PaddleX integrates image classification, target detection, semantic segmentation, and instance segmentation task capabilities in the field of Paddle Intelligent Vision. It connects the entire deep learning development process end-to-end from data preparation, model training and optimization to multi-end deployment, and provides a unified task API interface and Graphical development interface Demo. Developers do not need to install different packages separately, and can quickly complete the entire process of flying paddle development in a low-code form.

PaddleX has been verified by practical application scenarios in more than ten industries such as quality inspection, security, inspection, remote sensing, retail, and medical care. It has accumulated practical industry experience and provided rich case practice tutorials to help developers implement industrial practices throughout the process.

Install

PaddleX provides three development modes to meet the different needs of users

1. Python development model

Through the concise and easy-to-understand Python API, it gives developers the smoothest deep learning development experience while taking into account comprehensive functions, development flexibility, and integration convenience.

pre-dependency

paddlepaddle >= 1.8.4

python >= 3.6

cython

pycocotools

pip install paddlex -i https://mirror.baidu.com/pypi/simple

For detailed installation methods, please refer to PaddleX installation

2. Padlde GUI mode

The visual client developed without code is implemented using Paddle API, allowing developers to quickly verify industrial projects and providing a reference for users to develop their own deep learning software/applications.

Go to the PaddleX official website and apply to download the PaddleX GUI one-click green installation package.

Go to the PaddleX GUI tutorial to learn more about how to use the PaddleX GUI.

PaddleX GUI installation environment instructions

3. PaddleX Restful:

Use GUI and Web Demo developed based on RESTful API to realize remote deep learning full process development; at the same time, developers can also develop personalized visual interface based on RESTful API

Go to PaddleX RESTful API usage tutorial

PaddleX update log

v2.0.0.rc0

Full support for Flying Paddle 2.0 dynamic graphics, an easier-to-use development model

PP-YOLOv2 was added to the target detection task, the COCO test data set accuracy reached 49.5%, and the V100 prediction speed reached 68.9 FPS

A new 4.2MB ultra-lightweight model PP-YOLO tiny is added to the target detection task.

A new real-time segmentation model BiSeNetV2 is added to the semantic segmentation task.

C++ deployment module fully upgraded

PaddleInference deployment adaptation 2.0 prediction library

Supports model deployment of PaddleDetection, PaddleSeg, PaddleClas and PaddleX

Added GPU multi-card prediction based on PaddleInference

GPU deployment adds ONNX-based TensorRT high-performance acceleration engine deployment method

GPU deployment adds Triton service-based deployment method based on ONNX