Skip to content

Latest commit

 

History

History
133 lines (101 loc) · 6.37 KB

README_en.md

File metadata and controls

133 lines (101 loc) · 6.37 KB
Raw

简体中文 | English

YOLO_ZOO

YOLO_ZOO is an end-to-end object detector development kit based on the pytorch framework to reproduce the latest Yolo algorithm. It aims to help developers complete the whole process of detection model training, accuracy and speed optimization, and deployment faster and better. YOLO_ZOO implements a variety of mainstream YOLO target detection algorithms with a modular design, and provides a wealth of modules such as data enhancement, network components, and loss functions.

Currently, all models in the detection library require the use of PyTorch 1.5 and above or the appropriate develop version.

Content

Introduction

Features:

  • Rich models:

    YOLO_ZOO provides a wealth of models, covering the reproduction of the latest YOLO detection algorithms, include YOLO series target detection algorithms such as YOLOv5, YOLOv4, PP-YOLO, YOLOv3, etc.

  • High flexibility:

    YOLO_ZOO decouples various components through modular design, and can easily build various detection models based on configuration files.

Supported models:

More Backone:

  • DarkNet
  • CSPDarkNet
  • ResNet
  • YOLOv5Darknet

Data enhancement method:

  • Mosaic
  • MixUp
  • Resize
  • LetterBox
  • RandomCrop
  • RandomFlip
  • RandomHSV
  • RandomBlur
  • RandomNoise
  • RandomAffine
  • RandomTranslation
  • Normalize
  • ImageToTensor
  • Please refer to [here] for related configuration instructions

Loss function support:

  • bbox loss (IOU,GIOU,DIOU,CIOU)
  • confidence loss (YOLOv4, YOLOv5, PP-YOLO)
  • IOU_Aware_Loss(PP-YOLO)
  • FocalLoss

Training skills support:

  • Exponential Moving Average
  • Warm up
  • Gradient shear
  • Gradient cumulative update
  • Multi-scale training
  • Learning rate adjustment: Fixed, Step, Exp, Poly, Inv, Consine
  • Label Smooth
  • Strong explanation Through experimental comparison, it is found that YOLOv5's positive and negative sample division definition and loss function definition make the model converge faster, far exceeding the original yolo series' division and loss definition of positive and negative samples. If the card resources are insufficient and you want to converge the model in a short time, you can use yolov5's positive and negative sample partition and loss function definition, the relevant parameter is yolo_loss_type=yolov5.
  • Additional supplement YOLOv5's definition of positive samples: As long as the ratio of the true frame to the given anchor frame is within 4 times at different scales, the anchor frame can be responsible for predicting the true value frame. And according to the offset of gx and gy in the center of the grid, two additional grid coordinates will be added to predict. Through this series of operations, the number of positive samples is increased and the model convergence speed is accelerated. For the original YOLO series, for the true frame, only the anchor frame with the largest IOU intersection at this scale is responsible for predicting the true frame at different scales, and other anchor frames are not responsible. Therefore, due to the smaller positive sample size, the model converges faster. slow.

Extended features:

  • Group Norm
  • Modulated Deformable Convolution
  • Focus
  • Spatial Pyramid Pooling
  • FPN-PAN
  • coord conv
  • drop block
  • SAM

Code structure description

yolo_zoo
├──cfg              #The directory where the model configuration file is located (yolov5, yolov4, etc.)
├──tools            #toolkit, contains training code, test code and inference code entry.
├──yolodet          #YOLO detection framework core code base
│ ├──apis           #Provide an interface for training, testing and inference of the detection framework and model preservation
│ ├──dataset        #Contains general methods such as DateSet, DateLoader and data enhancement
│ ├──models         #The core components of the YOLO detection framework are assembled
│ │ ├──detectors    #Assembly of all types of detectors
│ │ ├──backbones    #All backbone network gathering places
│ │ ├──necks        #All necks gathering place
│ │ ├──heads        #heads assembly place
│ │ ├──loss         #The gathering place of all loss functions
│ │ ├──hooks        #hooks assembly place (learning rate adjustment, model saving, training log, weight update, etc.)
│ │ ├──utils        #All tools and methods gathering place

Installation

Please refer to INSTALL.md for installation and data set preparation.

Quick start

Please refer to GETTING_STARTED.md for basic usage of YOLODet.

Pretraining Model

Please click 【here】 to view the pre-training model.

Important

Due to the testing framework for personal leisure, in the love of deep learning, written by herself, but also because their cash-strapped, do not have enough graphics resources, provide the MSCOCO large data sets of training model for the complete training model, behind will provide a complete version of the training model of the hall, please look forward to.For the model with small data sets that I have tested and verified and that I have used the framework to train in the actual project, there is no problem, and the accuracy and speed can be guaranteed.

Thanks