diff --git a/README_cn.md b/README_cn.md index c4f19520815e7ce683d85b905582c1b1b4b97f12..c6f1c12aa1cc99d55b56ffe51de560b025ef6ac6 100644 --- a/README_cn.md +++ b/README_cn.md @@ -24,20 +24,18 @@ ## 产品动态 -- 🔥 **2022.8.09：[YOLO家族全系列模型](https://github.com/nemonameless/PaddleDetection_YOLOSeries)发布** - - 全面覆盖的YOLO家族经典与最新模型: 包括YOLOv3，百度飞桨自研的实时高精度目标检测检测模型PP-YOLOE，以及前沿检测算法YOLOv4、YOLOv5、YOLOX，MT-YOLOv6及YOLOv7 - - 更强的模型性能：基于各家前沿YOLO算法进行创新并升级，缩短训练周期5~8倍，精度普遍提升1%~5% mAP；使用模型压缩策略实现精度无损的同时速度提升30%以上 - - 完备的端到端开发支持：支持从模型训练、评估、预测到模型量化压缩，部署多种硬件的端到端开发全流程。同时支持不同模型算法灵活切换，一键实现算法二次开发 - -- 🔥 **2022.8.01：发布[PP-TinyPose升级版](./configs/keypoint/tiny_pose/). 在健身、舞蹈等场景的业务数据集端到端AP提升9.1** - - 新增体育场景真实数据，复杂动作识别效果显著提升，覆盖侧身、卧躺、跳跃、高抬腿等非常规动作 - - 检测模型采用[PP-PicoDet增强版](./configs/picodet/README.md)，在COCO数据集上精度提升3.1% - - 关键点稳定性增强，新增滤波稳定方式，使得视频预测结果更加稳定平滑 - -- 2022.7.14：[行人分析工具PP-Human v2](./deploy/pipeline)发布 - - 四大产业特色功能：高性能易扩展的五大复杂行为识别、闪电级人体属性识别、一行代码即可实现的人流检测与轨迹留存以及高精度跨镜跟踪 - - 底层核心算法性能强劲：覆盖行人检测、跟踪、属性三类核心算法能力，对目标人数、光线、背景均无限制 - - 极低使用门槛：提供保姆级全流程开发及模型优化策略、一行命令完成推理、兼容各类数据输入格式 +- 🔥 **2022.8.26：PaddleDetection发布[release/2.5版本](https://github.com/PaddlePaddle/PaddleDetection/tree/release/2.5)** + - 🗳 特色模型： + - 发布[PP-YOLOE+](configs/ppyoloe)，最高精度提升2.4% mAP，达到54.9% mAP，模型训练收敛速度提升3.75倍，端到端预测速度最高提升2.3倍；多个下游任务泛化性提升 + - 发布[PicoDet-NPU](configs/picodet)模型，支持模型全量化部署；新增[PicoDet](configs/picodet)版面分析模型 + - 发布[PP-TinyPose升级版](./configs/keypoint/tiny_pose/)增强版，在健身、舞蹈等场景精度提升9.1% AP，支持侧身、卧躺、跳跃、高抬腿等非常规动作 + - 🔮 场景能力： + - 发布行人分析工具[PP-Human v2](./deploy/pipeline)，新增打架、打电话、抽烟、闯入四大行为识别，底层算法性能升级，覆盖行人检测、跟踪、属性三类核心算法能力，提供保姆级全流程开发及模型优化策略，支持在线视频流输入 + - 首次发布[PP-Vehicle](./deploy/pipeline)，提供车牌识别、车辆属性分析（颜色、车型）、车流量统计以及违章检测四大功能，兼容图片、在线视频流、视频输入，提供完善的二次开发文档教程 + - 💡 前沿算法： + - 全面覆盖的[YOLO家族](https://github.com/nemonameless/PaddleDetection_YOLOSeries)经典与最新模型: 包括YOLOv3，百度飞桨自研的实时高精度目标检测检测模型PP-YOLOE，以及前沿检测算法YOLOv4、YOLOv5、YOLOX，MT-YOLOv6及YOLOv7 + - 新增基于[ViT](configs/vitdet)骨干网络高精度检测模型，COCO数据集精度达到55.7% mAP；新增[OC-SORT](configs/mot/ocsort)多目标跟踪模型；新增[ConvNeXt](configs/convnext)骨干网络 + - 📋 产业范例：新增[智能健身](https://aistudio.baidu.com/aistudio/projectdetail/4385813)、[打架识别](https://aistudio.baidu.com/aistudio/projectdetail/4086987?channelType=0&channel=0)、[来客分析](https://aistudio.baidu.com/aistudio/projectdetail/4230123?channelType=0&channel=0)、车辆结构化范例 - 2022.3.24：PaddleDetection发布[release/2.4版本](https://github.com/PaddlePaddle/PaddleDetection/tree/release/2.4) - 发布高精度云边一体SOTA目标检测模型[PP-YOLOE](configs/ppyoloe)，提供s/m/l/x版本，l版本COCO test2017数据集精度51.6%，V100预测速度78.1 FPS，支持混合精度训练，训练较PP-YOLOv2加速33%，全系列多尺度模型，满足不同硬件算力需求，可适配服务器、边缘端GPU及其他服务器端AI加速卡。 @@ -63,7 +61,7 @@ - **高性能**: 基于飞桨的高性能内核，模型训练速度及显存占用优势明显。支持FP16训练, 支持多机训练。
- +
## 技术交流 @@ -112,6 +110,7 @@
  • PP-YOLOv1/v2
  • PP-YOLO-Tiny
  • PP-YOLOE
    • +
  • PP-YOLOE+
  • YOLOX
  • SSD
  • CenterNet
    • @@ -141,6 +140,7 @@
  • FairMOT
  • DeepSORT
  • ByteTrack
    • +
  • OC-SORT
    • KeyPoint-Detection
    @@ -259,11 +261,10 @@ **说明：** -- `CBResNet`为`Cascade-Faster-RCNN-CBResNet200vd-FPN`模型，COCO数据集mAP高达53.3% +- `ViT`为`ViT-Cascade-Faster-RCNN`模型，COCO数据集mAP高达55.7% - `Cascade-Faster-RCNN`为`Cascade-Faster-RCNN-ResNet50vd-DCN`，PaddleDetection将其优化到COCO数据mAP为47.8%时推理速度为20FPS -- `PP-YOLO`在COCO数据集精度45.9%，Tesla V100预测速度72.9FPS，精度速度均优于[YOLOv4](https://arxiv.org/abs/2004.10934) -- `PP-YOLO v2`是对`PP-YOLO`模型的进一步优化，在COCO数据集精度49.5%，Tesla V100预测速度68.9FPS - `PP-YOLOE`是对`PP-YOLO v2`模型的进一步优化，在COCO数据集精度51.6%，Tesla V100预测速度78.1FPS +- `PP-YOLOE+`是对`PPOLOE`模型的进一步优化，在COCO数据集精度53.3%，Tesla V100预测速度78.1FPS - [`YOLOX`](configs/yolox)和[`YOLOv5`](https://github.com/nemonameless/PaddleDetection_YOLOSeries/tree/develop/configs/yolov5)均为基于PaddleDetection复现算法 - 图中模型均可在[模型库](#模型库)中获取 @@ -290,14 +291,14 @@
    1. 通用检测 -#### [PP-YOLOE](./configs/ppyoloe)系列 推荐场景：Nvidia V100, T4等云端GPU和Jetson系列等边缘端设备 +#### [PP-YOLOE+](./configs/ppyoloe)系列 推荐场景：Nvidia V100, T4等云端GPU和Jetson系列等边缘端设备 | 模型名称 | COCO精度（mAP） | V100 TensorRT FP16速度(FPS) | 配置文件 | 模型下载 | |:---------- |:-----------:|:-------------------------:|:-----------------------------------------------------:|:------------------------------------------------------------------------------------:| -| PP-YOLOE-s | 42.7 | 333.3 | [链接](configs/ppyoloe/ppyoloe_crn_l_300e_coco.yml) | [下载地址](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_l_300e_coco.pdparams) | -| PP-YOLOE-m | 48.6 | 208.3 | [链接](configs/ppyolo/ppyolo_r50vd_dcn_2x_coco.yml) | [下载地址](https://paddledet.bj.bcebos.com/models/ppyolo_r50vd_dcn_2x_coco.pdparams) | -| PP-YOLOE-l | 50.9 | 149.2 | [链接](configs/ppyolo/ppyolov2_r50vd_dcn_365e_coco.yml) | [下载地址](https://paddledet.bj.bcebos.com/models/ppyolov2_r50vd_dcn_365e_coco.pdparams) | -| PP-YOLOE-x | 51.9 | 95.2 | [链接](configs/ppyolo/ppyolov2_r50vd_dcn_365e_coco.yml) | [下载地址](https://paddledet.bj.bcebos.com/models/ppyolov2_r50vd_dcn_365e_coco.pdparams) | +| PP-YOLOE+_s | 43.9 | 333.3 | [链接](configs/ppyoloe/ppyoloe_plus_crn_s_80e_coco.yml) | [下载地址](https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_s_80e_coco.pdparams) | +| PP-YOLOE+_m | 50.0 | 208.3 | [链接](configs/ppyoloe/ppyoloe_plus_crn_m_80e_coco.yml) | [下载地址](https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_m_80e_coco.pdparams) | +| PP-YOLOE+_l | 53.3 | 149.2 | [链接](configs/ppyoloe/ppyoloe_plus_crn_l_80e_coco.yml) | [下载地址](https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_m_80e_coco.pdparams) | +| PP-YOLOE+_x | 54.9 | 95.2 | [链接](configs/ppyoloe/ppyoloe_plus_crn_x_80e_coco.yml) | [下载地址](https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_x_80e_coco.pdparams) | #### [PP-PicoDet](./configs/picodet)系列 推荐场景：ARM CPU(RK3399, 树莓派等) 和NPU(比特大陆，晶晨等)移动端芯片和x86 CPU设备 @@ -354,6 +355,7 @@ | ByteTrack | SDE多目标跟踪算法 仅包含检测模型 | 云边端 | MOT-17 half val: 77.3 | [链接](configs/mot/bytetrack/detector/yolox_x_24e_800x1440_mix_det.yml) | [下载地址](https://paddledet.bj.bcebos.com/models/mot/deepsort/yolox_x_24e_800x1440_mix_det.pdparams) | | JDE | JDE多目标跟踪算法 多任务联合学习方法 | 云边端 | MOT-16 test: 64.6 | [链接](configs/mot/jde/jde_darknet53_30e_1088x608.yml) | [下载地址](https://paddledet.bj.bcebos.com/models/mot/jde_darknet53_30e_1088x608.pdparams) | | FairMOT | JDE多目标跟踪算法 多任务联合学习方法 | 云边端 | MOT-16 test: 75.0 | [链接](configs/mot/fairmot/fairmot_dla34_30e_1088x608.yml) | [下载地址](https://paddledet.bj.bcebos.com/models/mot/fairmot_dla34_30e_1088x608.pdparams) | +| OC-SORT | SDE多目标跟踪算法 仅包含检测模型 | 云边端 | MOT-17 half val: 75.5 | [链接](configs/mot/ocsort/ocsort_yolox.yml) | - | #### 其他多目标跟踪模型 [文档链接](configs/mot) diff --git a/README_en.md b/README_en.md index af9948f40f5c2dbbda4231344c4e7ac3c6dae58b..aae98b999b94a72a1a20017dfcbec6ba5a1c9686 100644 --- a/README_en.md +++ b/README_en.md @@ -23,21 +23,25 @@ ## Product Update -- 🔥 **2022.8.09：Release [YOLO series model zoo](https://github.com/nemonameless/PaddleDetection_YOLOSeries)** - - Comprehensive coverage of classic and latest models of the YOLO series: Including YOLOv3，Paddle real-time object detection model PP-YOLOE, and frontier detection algorithms YOLOv4, YOLOv5, YOLOX, MT-YOLOv6 and YOLOv7 - - Better model performance：Upgrade based on various YOLO algorithms, shorten training time in 5-8 times and the accuracy is generally improved by 1%-5% mAP. The model compression strategy is used to achieve 30% improvement in speed without precision loss - - Complete end-to-end development support：End-to-end development pipieline including training, evaluation, inference, model compression and deployment on various hardware. Meanwhile, support flexible algorithnm switch and implement customized development efficiently - -- 🔥 **2022.8.01：Release [PP-TinyPose plus](./configs/keypoint/tiny_pose/). The end-to-end precision improves 9.1% AP in dataset - of fitness and dance scenes** - - Increase data of sports scenes, and the recognition performance of complex actions is significantly improved, covering actions such as sideways, lying down, jumping, and raising legs - - Detection model uses PP-PicoDet plus and the precision on COCO dataset is improved by 3.1% mAP - - The stability of keypoints is enhanced. Implement the filter stabilization method to make the video prediction result more stable and smooth. - -- 2022.7.14：Release [pedestrian analysis tool PP-Human v2](./deploy/pipeline) - - Four major functions: five complicated action recognition with high performance and Flexible, real-time human attribute recognition, visitor flow statistics and high-accuracy multi-camera tracking. - - High performance algorithm: including pedestrian detection, tracking, attribute recognition which is robust to the number of targets and the variant of background and light. - - Highly Flexible: providing complete introduction of end-to-end development and optimization strategy, simple command for deployment and compatibility with different input format. +- 🔥 **2022.8.26：PaddleDetection releases[release/2.5 version](https://github.com/PaddlePaddle/PaddleDetection/tree/release/2.5)** + + - 🗳 Model features： + + - Release [PP-YOLOE+](configs/ppyoloe): Increased accuracy by a maximum of 2.4% mAP to 54.9% mAP, 3.75 times faster model training convergence rate, and up to 2.3 times faster end-to-end inference speed; improved generalization for multiple downstream tasks + - Release [PicoDet-NPU](configs/picodet) model which supports full quantization deployment of models; add [PicoDet](configs/picodet) layout analysis model + - Release [PP-TinyPose Plus](./configs/keypoint/tiny_pose/). With 9.1% AP accuracy improvement in physical exercise, dance, and other scenarios, our PP-TinyPose Plus supports unconventional movements such as turning to one side, lying down, jumping, and high lifts + + - 🔮 Functions in different scenarios + + - Release the pedestrian analysis tool [PP-Human v2](./deploy/pipeline). It introduces four new behavior recognition: fighting, telephoning, smoking, and trespassing. The underlying algorithm performance is optimized, covering three core algorithm capabilities: detection, tracking, and attributes of pedestrians. Our model provides end-to-end development and model optimization strategies for beginners and supports online video streaming input. + - First release [PP-Vehicle](./deploy/pipeline), which has four major functions: license plate recognition, vehicle attribute analysis (color, model), traffic flow statistics, and violation detection. It is compatible with input formats, including pictures, online video streaming, and video. And we also offer our users a comprehensive set of tutorials for customization. + + - 💡 Cutting-edge algorithms： + + - Covers [YOLO family](https://github.com/nemonameless/PaddleDetection_YOLOSeries) classic and latest models: YOLOv3, PP-YOLOE (a real-time high-precision object detection model developed by Baidu PaddlePaddle), and cutting-edge detection algorithms such as YOLOv4, YOLOv5, YOLOX, MT-YOLOv6, and YOLOv7 + - Newly add high precision detection model based on [ViT](configs/vitdet) backbone network, with a 55.7% mAP accuracy on COCO dataset; newly add multi-object tracking model [OC-SORT](configs/mot/ocsort); newly add [ConvNeXt](configs/convnext) backbone network. + + - 📋 Industrial applications: Newly add [Smart Fitness](https://aistudio.baidu.com/aistudio/projectdetail/4385813), [Fighting recognition](https://aistudio.baidu.com/aistudio/projectdetail/4086987?channelType=0&channel=0),[ and Visitor Analysis](https://aistudio.baidu.com/aistudio/projectdetail/4230123?channelType=0&channel=0). - 2022.3.24：PaddleDetection released[release/2.4 version](https://github.com/PaddlePaddle/PaddleDetection/tree/release/2.4) - Release high-performanace SOTA object detection model [PP-YOLOE](configs/ppyoloe). It integrates cloud and edge devices and provides S/M/L/X versions. In particular, Verson L has the accuracy as 51.4% on COCO test 2017 dataset, inference speed as 78.1 FPS on a single Test V100. It supports mixed precision training, 33% faster than PP-YOLOv2. Its full range of multi-sized models can meet different hardware arithmetic requirements, and adaptable to server, edge-device GPU and other AI accelerator cards on servers. @@ -65,7 +69,7 @@ - **High Performance**: Due to the high performance core, PaddlePaddle has clear advantages in training speed and memory occupation. It also supports FP16 training and multi-machine training.
    - newstructure + newstructure
    Exchanges @@ -111,6 +115,7 @@
  • PP-YOLOv1/v2
  • PP-YOLO-Tiny
  • PP-YOLOE
    • +
  • PP-YOLOE+
  • YOLOX
  • SSD
  • CenterNet
    • @@ -140,6 +145,7 @@
  • FairMOT
  • DeepSORT
  • ByteTrack
    • +
  • OC-SORT
    • KeyPoint-Detection
    @@ -258,11 +266,10 @@ The comparison between COCO mAP and FPS on Tesla V100 of representative models o **Clarification：** -- `CBResNet` stands for `Cascade-Faster-RCNN-CBResNet200vd-FPN`, which has highest mAP on COCO as 53.3% +- `ViT` stands for `ViT-Cascade-Faster-RCNN`, which has highest mAP on COCO as 55.7% - `Cascade-Faster-RCNN`stands for `Cascade-Faster-RCNN-ResNet50vd-DCN`, which has been optimized to 20 FPS inference speed when COCO mAP as 47.8% in PaddleDetection models -- `PP-YOLO` reached accuracy as 45.9% on COCO dataset, inference speed as 72.9 FPS on Tesla V100, higher than [YOLOv4]([[2004.10934] YOLOv4: Optimal Speed and Accuracy of Object Detection](https://arxiv.org/abs/2004.10934)) in terms of speed and accuracy -- `PP-YOLO v2`are optimized `PP-YOLO`. It reached accuracy as 49.5% on COCO dataset, inference speed as 68.9 FPS on Tesla V100. -- `PP-YOLOE`are optimized `PP-YOLO v2`. It reached accuracy as 51.4% on COCO dataset, inference speed as 78.1 FPS on Tesla V100 +- `PP-YOLOE` are optimized `PP-YOLO v2`. It reached accuracy as 51.4% on COCO dataset, inference speed as 78.1 FPS on Tesla V100 +- `PP-YOLOE+` are optimized `PP-YOLOE`. It reached accuracy as 53.3% on COCO dataset, inference speed as 78.1 FPS on Tesla V100 - The models in the figure are available in the[ model library](#模型库) @@ -292,10 +299,10 @@ The comparison between COCO mAP and FPS on Qualcomm Snapdragon 865 processor of | Model | COCO Accuracy（mAP） | V100 TensorRT FP16 Speed(FPS) | Configuration | Download | |:---------- |:------------------:|:-----------------------------:|:-------------------------------------------------------:|:----------------------------------------------------------------------------------------:| -| PP-YOLOE-s | 42.7 | 333.3 | [Link](configs/ppyoloe/ppyoloe_crn_l_300e_coco.yml) | [Download](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_l_300e_coco.pdparams) | -| PP-YOLOE-m | 48.6 | 208.3 | [Link](configs/ppyolo/ppyolo_r50vd_dcn_2x_coco.yml) | [Download](https://paddledet.bj.bcebos.com/models/ppyolo_r50vd_dcn_2x_coco.pdparams) | -| PP-YOLOE-l | 50.9 | 149.2 | [Link](configs/ppyolo/ppyolov2_r50vd_dcn_365e_coco.yml) | [Download](https://paddledet.bj.bcebos.com/models/ppyolov2_r50vd_dcn_365e_coco.pdparams) | -| PP-YOLOE-x | 51.9 | 95.2 | [Link](configs/ppyolo/ppyolov2_r50vd_dcn_365e_coco.yml) | [Download](https://paddledet.bj.bcebos.com/models/ppyolov2_r50vd_dcn_365e_coco.pdparams) | +| PP-YOLOE+_s | 43.9 | 333.3 | [link](configs/ppyoloe/ppyoloe_plus_crn_s_80e_coco.yml) | [download](https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_s_80e_coco.pdparams) | +| PP-YOLOE+_m | 50.0 | 208.3 | [link](configs/ppyoloe/ppyoloe_plus_crn_m_80e_coco.yml) | [download](https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_m_80e_coco.pdparams) | +| PP-YOLOE+_l | 53.3 | 149.2 | [link](configs/ppyoloe/ppyoloe_plus_crn_l_80e_coco.yml) | [download](https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_m_80e_coco.pdparams) | +| PP-YOLOE+_x | 54.9 | 95.2 | [link](configs/ppyoloe/ppyoloe_plus_crn_x_80e_coco.yml) | [download](https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_x_80e_coco.pdparams) | #### PP-PicoDet series Recommended scenarios: Mobile chips and x86 CPU devices, such as ARM CPU(RK3399, Raspberry Pi) and NPU(BITMAIN) @@ -351,6 +358,7 @@ The comparison between COCO mAP and FPS on Qualcomm Snapdragon 865 processor of | ByteTrack | SDE Multi-object tracking algorithm with detection model only | Edge-Cloud end | MOT-17 half val: 77.3 | [Link](configs/mot/bytetrack/detector/yolox_x_24e_800x1440_mix_det.yml) | [Download](https://paddledet.bj.bcebos.com/models/mot/deepsort/yolox_x_24e_800x1440_mix_det.pdparams) | | JDE | JDE multi-object tracking algorithm multi-task learning | Edge-Cloud end | MOT-16 test: 64.6 | [Link](configs/mot/jde/jde_darknet53_30e_1088x608.yml) | [Download](https://paddledet.bj.bcebos.com/models/mot/jde_darknet53_30e_1088x608.pdparams) | | FairMOT | JDE multi-object tracking algorithm multi-task learning | Edge-Cloud end | MOT-16 test: 75.0 | [Link](configs/mot/fairmot/fairmot_dla34_30e_1088x608.yml) | [Download](https://paddledet.bj.bcebos.com/models/mot/fairmot_dla34_30e_1088x608.pdparams) | +| OC-SORT | SDE multi-object tracking algorithm with detection model only | Edge-Cloud end | MOT-16 half val: 75.5 | [Link](configs/mot/ocsort/ocsort_yolox.yml) | - | #### Other multi-object tracking models [docs](configs/mot) diff --git a/docs/CHANGELOG.md b/docs/CHANGELOG.md index e19e6867a15ea18ddb8f85f46f6e020b79d4ebf6..f5fef04f1c946f69bda11ee16d7222c0dc6e8eab 100644 --- a/docs/CHANGELOG.md +++ b/docs/CHANGELOG.md @@ -4,6 +4,63 @@ ## 最新版本信息 +### 2.5(08.26/2022) + +- 特色模型 + - PP-YOLOE+： + - 发布PP-YOLOE+模型，COCO test2017数据集精度提升0.7%-2.4% mAP，模型训练收敛速度提升3.75倍，端到端预测速度提升1.73-2.3倍 + - 发布智慧农业，夜间安防检测，工业质检场景预训练模型，精度提升1.3%-8.1% mAP + - 支持分布式训练、在线量化、serving部署等10大高性能训练部署能力，新增C++/Python Serving、TRT原生推理、ONNX Runtime等5+部署demo教程 + - PP-PicoDet： + - 发布PicoDet-NPU模型，支持模型全量化部署 + - 新增PicoDet版面分析模型，基于FGD蒸馏算法精度提升0.5% mAP + - PP-TinyPose + - 发布PP-TinyPose增强版，在健身、舞蹈等场景的业务数据集端到端AP提升9.1% AP + - 覆盖侧身、卧躺、跳跃、高抬腿等非常规动作 + - 新增滤波稳定模块，关键点稳定性显著增强 + +- 场景能力 + - PP-Human v2 + - 发布PP-Human v2，支持四大产业特色功能：多方案行为识别案例库、人体属性识别、人流检测与轨迹留存以及高精度跨镜跟踪 + - 底层算法能力升级，行人检测精度提升1.5% mAP；行人跟踪精度提升10.2% MOTA，轻量级模型速度提升34%；属性识别精度提升0.6% ma，轻量级模型速度提升62.5% + - 提供全流程教程，覆盖数据采集标注，模型训练优化和预测部署，及pipeline中后处理代码修改 + - 新增在线视频流输入支持 + - 易用性提升，一行代码执行功能，执行流程判断、模型下载背后自动完成。 + - PP-Vehicle + - 全新发布PP-Vehicle，支持四大交通场景核心功能：车牌识别、属性识别、车流量统计、违章检测 + - 车牌识别支持基于PP-OCR v3的轻量级车牌识别模型 + - 车辆属性识别支持基于PP-LCNet多标签分类模型 + - 兼容图片、视频、在线视频流等各类数据输入格式 + - 易用性提升，一行代码执行功能，执行流程判断、模型下载背后自动完成。 + +- 前沿算法 + - YOLO家族全系列模型 + - 发布YOLO家族全系列模型，覆盖前沿检测算法YOLOv5、MT-YOLOv6及YOLOv7 + - 基于ConvNext骨干网络，YOLO各算法训练周期缩5-8倍，精度普遍提升1%-5% mAP；使用模型压缩策略实现精度无损的同时速度提升30%以上 + - 新增基于ViT骨干网络高精度检测模型，COCO数据集精度达到55.7% mAP + - 新增OC-SORT多目标跟踪模型 + - 新增ConvNeXt骨干网络 + +- 产业实践范例教程 + - 基于PP-TinyPose增强版的智能健身动作识别 + - 基于PP-Human的打架识别 + - 基于PP-Human的营业厅来客分析 + - 基于PP-Vehicle的车辆结构化分析 + - 基于PP-YOLOE+的PCB电路板缺陷检测 + +- 框架能力 + - 功能新增 + - 新增自动压缩工具支持并提供demo，PP-YOLOE l版本精度损失0.3% mAP，V100速度提升13% + - 新增PaddleServing python/C++和ONNXRuntime部署demo + - 新增PP-YOLOE 端到端TensorRT部署demo + - 新增FGC蒸馏算法，RetinaNet精度提升3.3% + - 新增分布式训练文档 + - 功能完善/Bug修复 + - 修复Windows c++部署编译问题 + - 修复VOC格式数据预测时保存结果问题 + - 修复FairMOT c++部署检测框输出 + - 旋转框检测模型S2ANet支持batch size>1部署 + ### 2.4(03.24/2022) - PP-YOLOE： diff --git a/docs/CHANGELOG_en.md b/docs/CHANGELOG_en.md index a7e6d422611eae7f0cfa66deb56f8e53e493d8c2..601bb2c5b027bb100eb55aef34aebf221cd555dc 100644 --- a/docs/CHANGELOG_en.md +++ b/docs/CHANGELOG_en.md @@ -4,6 +4,68 @@ English | [简体中文](./CHANGELOG.md) ## Last Version Information +### 2.5(08.26/2022) + +- Featured model + + - PP-YOLOE+： + - Released PP-YOLOE+ model, with a 0.7%-2.4% mAP improvement on COCO test2017. 3.75 times faster model training convergence rate and 1.73-2.3 times faster end-to-end inference speed + - Released pre-trained models for smart agriculture, night security detection, and industrial quality inspection with 1.3%-8.1% mAP accuracy improvement + - supports 10 high-performance training deployment capabilities, including distributed training, online quantization, and serving deployment. We also provide more than five new deployment demos, such as C++/Python Serving, TRT native inference, and ONNX Runtime + - PP-PicoDet： + - Release the PicoDet-NPU model to support full quantization of model deployment + - Add PicoDet layout analysis model with 0.5% mAP accuracy improvement due to FGD distillation algorithm + - PP-TinyPose + - Release PP-TinyPose Plus with 9.1% end-to-end AP improvement for business data sets such as physical exercises, dance, and other scenarios + - Covers unconventional movements such as turning to one side, lying down, jumping, high lift + - Add stabilization module (via filter) to significantly improve the stability at key points + +- Functions in different scenarios + + - PP-Human v2 + - Release PP-Human v2, which supports four industrial features: behavioral recognition case zoo for multiple solutions, human attribute recognition, human traffic detection and trajectory retention, as well as high precision multi-camera tracking + - Upgraded underlying algorithm capabilities: 1.5% mAP improvement in pedestrian detection accuracy; 10.2% MOTA improvement in pedestrian tracking accuracy, 34% speed improvement in the lightweight model; 0.6% ma improvement in attribute recognition accuracy, 62.5% speed improvement in the lightweight model + - Provides comprehensive tutorials covering data collection and annotation, model training optimization and prediction deployment, and post-processing code modification in the pipeline + - Supports online video streaming input + - Become more user-friendly with a one-line code execution function that automates the process determination and model download + - PP-Vehicle + - Launch PP-Vehicle, which supports four core functions for traffic application: license plate recognition, attribute recognition, traffic flow statistics, and violation detection + - License plate recognition supports a lightweight model based on PP-OCR v3 + - Vehicle attribute recognition supports a multi-label classification model based on PP-LCNet + - Compatible with various data input formats such as pictures, videos and online video streaming + - Become more user-friendly with a one-line code execution function that automates the process determination and model download + +- Cutting-edge algorithms + + - YOLO Family + - Release the full range of YOLO family models covering the cutting-edge detection algorithms YOLOv5, MT-YOLOv6 and YOLOv7 + - Based on the ConvNext backbone network, YOLO's algorithm training periods are reduced by 5-8 times with accuracy generally improving by 1%-5% mAP; Thanks to the model compression strategy, its speed increased by over 30% with no loss of precision. + - Newly add high precision detection model based on [ViT](configs/vitdet) backbone network, with a 55.7% mAP accuracy on the COCO dataset + - Newly add multi-object tracking model [OC-SORT](configs/mot/ocsort) + - Newly add [ConvNeXt](configs/convnext) backbone network. + +- Industrial application + + - Intelligent physical exercise recognition based on PP-TinyPose Plus + - Fighting recognition based on PP-Human + - Business hall visitor analysis based on PP-Human + - Vehicle structuring analysis based on PP-Vehicle + - PCB board defect detection based on PP-YOLOE+ + +- Framework capabilities + + - New functions + - Release auto-compression tools and demos, 0.3% mAP accuracy loss for PP-YOLOE l version, while 13% speed increase for V100 + - Release PaddleServing python/C++ and ONNXRuntime deployment demos + - Release PP-YOLOE end-to-end TensorRT deployment demo + - Release FGC distillation algorithm with RetinaNet accuracy improved by 3.3% + - Release distributed training documentation + - Improvement and fixes + - Fix compilation problem with Windows c++ deployment + - Fix problems when saving results of inference data in VOC format + - Fix the detection box output of FairMOT c++ deployment + - Rotating frame detection model S2ANet supports batch size>1 deployment + ### 2.4(03.24/2022) - PP-YOLOE： diff --git a/docs/images/fps_map.png b/docs/images/fps_map.png index 5c22d725b01d374de8f394096f140b3d33cebfa7..0fbafcb4fb55fb3659a09b9ff20b6f82a9fe2ffc 100644 Binary files a/docs/images/fps_map.png and b/docs/images/fps_map.png differ diff --git a/docs/tutorials/INSTALL.md b/docs/tutorials/INSTALL.md index 418289b0d4b34c3e4b3df05adb52c0dc277e33dd..20018e891913c4ee373ee3279da2824b15c5d240 100644 --- a/docs/tutorials/INSTALL.md +++ b/docs/tutorials/INSTALL.md @@ -14,7 +14,7 @@ For general information about PaddleDetection, please see [README.md](https://gi - OS 64 bit - Python 3(3.5.1+/3.6/3.7/3.8/3.9)，64 bit - pip/pip3(9.0.1+), 64 bit -- CUDA >= 10.1 +- CUDA >= 10.2 - cuDNN >= 7.6 @@ -23,6 +23,7 @@ Dependency of PaddleDetection and PaddlePaddle: | PaddleDetection version | PaddlePaddle version | tips | | :----------------: | :---------------: | :-------: | | develop | >= 2.2.2 | Dygraph mode is set as default | +| release/2.5 | >= 2.2.2 | Dygraph mode is set as default | | release/2.4 | >= 2.2.2 | Dygraph mode is set as default | | release/2.3 | >= 2.2.0rc | Dygraph mode is set as default | | release/2.2 | >= 2.1.2 | Dygraph mode is set as default | @@ -40,11 +41,11 @@ Dependency of PaddleDetection and PaddlePaddle: ``` -# CUDA10.1 -python -m pip install paddlepaddle-gpu==2.2.0.post101 -f https://www.paddlepaddle.org.cn/whl/linux/mkl/avx/stable.html +# CUDA10.2 +python -m pip install paddlepaddle-gpu==2.2.2 -i https://mirror.baidu.com/pypi/simple # CPU -python -m pip install paddlepaddle -i https://mirror.baidu.com/pypi/simple +python -m pip install paddlepaddle==2.2.2 -i https://mirror.baidu.com/pypi/simple ``` - For more CUDA version or environment to quick install, please refer to the [PaddlePaddle Quick Installation document](https://www.paddlepaddle.org.cn/install/quick) @@ -116,7 +117,7 @@ OK We provide docker images containing the latest PaddleDetection code, and all environment and package dependencies are pre-installed. All you have to do is to **pull and run the docker image**. Then you can enjoy PaddleDetection without any extra steps. -Get these images and guidance in [docker hub](https://hub.docker.com/repository/docker/paddlecloud/paddledetection), including CPU, GPU, ROCm environment versions. +Get these images and guidance in [docker hub](https://hub.docker.com/repository/docker/paddlecloud/paddledetection), including CPU, GPU, ROCm environment versions. If you have some customized requirements about automatic building docker images, you can get it in github repo [PaddlePaddle/PaddleCloud](https://github.com/PaddlePaddle/PaddleCloud/tree/main/tekton). diff --git a/docs/tutorials/INSTALL_cn.md b/docs/tutorials/INSTALL_cn.md index 4afa7600f3dcd7ec1b3feb9d1cf0285b4de3bde5..0e13aeb05c5ef830053286f651b6f484dcf07d06 100644 --- a/docs/tutorials/INSTALL_cn.md +++ b/docs/tutorials/INSTALL_cn.md @@ -11,7 +11,7 @@ - OS 64位操作系统 - Python 3(3.5.1+/3.6/3.7/3.8/3.9)，64位版本 - pip/pip3(9.0.1+)，64位版本 -- CUDA >= 10.1 +- CUDA >= 10.2 - cuDNN >= 7.6 PaddleDetection 依赖 PaddlePaddle 版本关系： @@ -19,6 +19,7 @@ PaddleDetection 依赖 PaddlePaddle 版本关系： | PaddleDetection版本 | PaddlePaddle版本 | 备注 | | :------------------: | :---------------: | :-------: | | develop | >= 2.2.2 | 默认使用动态图模式 | +| release/2.5 | >= 2.2.2 | 默认使用动态图模式 | | release/2.4 | >= 2.2.2 | 默认使用动态图模式 | | release/2.3 | >= 2.2.0rc | 默认使用动态图模式 | | release/2.2 | >= 2.1.2 | 默认使用动态图模式 | @@ -34,11 +35,11 @@ PaddleDetection 依赖 PaddlePaddle 版本关系： ### 1. 安装PaddlePaddle ``` -# CUDA10.1 -python -m pip install paddlepaddle-gpu==2.2.0.post101 -f https://www.paddlepaddle.org.cn/whl/linux/mkl/avx/stable.html +# CUDA10.2 +python -m pip install paddlepaddle-gpu==2.2.2 -i https://mirror.baidu.com/pypi/simple # CPU -python -m pip install paddlepaddle -i https://mirror.baidu.com/pypi/simple +python -m pip install paddlepaddle==2.2.2 -i https://mirror.baidu.com/pypi/simple ``` - 更多CUDA版本或环境快速安装，请参考[PaddlePaddle快速安装文档](https://www.paddlepaddle.org.cn/install/quick) - 更多安装方式例如conda或源码编译安装方法，请参考[PaddlePaddle安装文档](https://www.paddlepaddle.org.cn/documentation/docs/zh/install/index_cn.html)