CDIoU and CDIoU loss is like a convenient plug-in that can be used in multiple models. CDIoU and CDIoU loss have different excellent performances in several models such as Faster R-CNN, YOLOv4, RetinaNet and . There is a maximum AP improvement of 1.9% and an average AP of 0.8% improvement on MS COCO dataset, compared to traditional evaluation-feedback modules. Here we just use as an example to illustrate the code.

Overview

CDIoU-CDIoUloss

CDIoU and CDIoU loss is like a convenient plug-in that can be used in multiple models. CDIoU and CDIoU loss have different excellent performances in several models such as Faster R-CNN, YOLOv4, RetinaNet and . There is a maximum AP improvement of 1.9% and an average AP of 0.8% improvement on MS COCO dataset, compared to traditional evaluation-feedback modules. Here we just use as an example to illustrate the code.

Control Distance IoU and Control Distance IoU Loss Function

by Chen Dong, Miao Duoqian

Introduction

Numerous improvements for feedback mechanisms have contributed to the great progress in object detection. In* this paper, we first present an evaluation-feedback module, which is proposed to consist of evaluation system and feedback mechanism. Then we analyze and summarize the disadvantages and improvements of traditional evaluation-feedback module. Finally, we focus on both the evaluation system and the feedback mechanism, and propose Control Distance IoU and Control Distance IoU loss function (or CDIoU and CDIoU loss for short) without increasing parameters or FLOPs in models, which show different significant enhancements on several classical and emerging models. Some experiments and comparative tests show that coordinated evaluation-feedback module can effectively improve model performance. CDIoU and CDIoU loss have different excellent performances in several models such as Faster R-CNN, YOLOv4, RetinaNet and ATSS. There is a maximum AP improvement of 1.9% and an average AP of 0.8% improvement on MS COCO dataset, compared to traditional evaluation-feedback modules.

There are some potential defects in the current mainstream target detection

  • It relies too much on the deepening of the backbone to extract features, so as to improve the accuracy of target detection;

  • The deepening of neural network, especially the deepening of backbone and neck, results in huge parameters and flops of the model;

  • Compared with the evaluation system (IoUs, the common ones are IoU and GIoU). At present, some new model optimization focuses more on the feedback mechanism (IoU losses), such as IoU loss, smooth loss, GIoU loss,CIoU loss, DIoU loss.

We propose Control Distance IoU and Control Distance IoU Loss Function (CDIoU and CDIoU loss for short).

Analysis of traditional IoUs and loss functions

  • Analysis of traditional IoUs

  • IoU: Smooth L1 Loss and IoU Loss

  • GIoU and GIoU Loss

  • DIoU loss and CIoU Loss

For more information, see Control Distance IoU and Control Distance IoU Loss Function for Better Bounding Box Regression

Installation

CDIoU and CDIoU loss is like a convenient plug-in that can be used in multiple models. CDIoU and CDIoU loss have different excellent performances in several models such as Faster R-CNN, YOLOv4, RetinaNet and ATSS. There is a maximum AP improvement of 1.9% and an average AP of 0.8% improvement on MS COCO dataset, compared to traditional evaluation-feedback modules. Here we just use ATSS as an example to illustrate the code.

These models use different frameworks, and some even have versions, so no code is provided in this article.


This ATSS implementation is based on FCOS and maskrcnn-benchmark and the installation is the same as them. Please check INSTALL.md for installation instructions.

ATSS bridges the gap between anchor-based and anchor-free detection via adaptive training sample selection. Comparison tests on ATSS exclude the essential interference between anchor-based and anchor-free detection. In these tests, the interference of positive and negative sample generation is eliminated, which give tests based on ATSS more representativeness.

CDIoU and CDIoU loss functions

CDIoU

Turning

For more information, see Control Distance IoU and Control Distance IoU Loss Function for Better Bounding Box Regression

Experiments

In order to verify the effectiveness of CDIoU and CDIoU loss in object detection, experiments are designed and applied to numerous models in this paper. These models encompass existing classical models and emerging models, reflecting certain robustness and wide adaptability.

image-20210315210941666

image-20210315214303929

For more information, see Control Distance IoU and Control Distance IoU Loss Function for Better Bounding Box Regression

Models

For your convenience, we provide the following trained models. All models are trained with 16 images in a mini-batch and frozen batch normalization (i.e., consistent with models in FCOS and maskrcnn_benchmark).

Model Multi-scale evaluation system feedback mechanism AP (val) AP (test-dev) pth
ATSS R 50 FPN 1x + CDIoU & loss NO CDIoU CDIoU loss 39.5 39.4 ATSS R 50 FPN 1x + CDIoU & loss
ATSS dcnv2 R 50 FPN 1x + CDIoU & loss NO CDIoU CDIoU loss 43.1 43.1 ATSS dcnv2 R 50 FPN 1x + CDIoU & loss
ATSS dcnv2 R 101 FPN 2x + CDIoU & loss NO CDIoU CDIoU loss 46.3 46.4 ATSS dcnv2 R 101 FPN 2x + CDIoU & loss
ATSS X 101 32x8d FPN 2x + CDIoU & loss NO CDIoU CDIoU loss 45.1 45.2 ATSS X 101 32x8d FPN 2x + CDIoU & loss
ATSS dcnv2 X 101 32x8d FPN 2x + CDIoU & loss NO CDIoU CDIoU loss 48.1 47.9 ATSS dcnv2 X 101 32x8d FPN 2x + CDIoU & loss
ATSS dcnv2 X 101 32x8d FPN 2x(MS) + CDIoU & loss YES CDIoU CDIoU loss 50.9 50.7 ATSS dcnv2 X 101 32x8d FPN 2x(MS) + CDIoU & loss

[1] The testing time is taken from FCOS, because our method only redefines positive and negative training samples without incurring any additional overhead. [2] 1x and 2x mean the model is trained for 90K and 180K iterations, respectively. [3] All results are obtained with a single model and without any test time data augmentation such as multi-scale, flipping and etc.. [4] dcnv2 denotes deformable convolutional networks v2. Note that for ResNet based models, we apply deformable convolutions from stage c3 to c5 in backbones. For ResNeXt based models, only stage c4 and c5 use deformable convolutions. All models use deformable convolutions in the last layer of detector towers. [5] The model ATSS_dcnv2_X_101_64x4d_FPN_2x with multi-scale testing achieves 50.7% in AP on COCO test-dev. Please use TEST.BBOX_AUG.ENABLED True to enable multi-scale testing.

MSCOCO test-dev

image-20210315222453294

Tips to improve performances

  • Floating learning rate

It is a consensus that the learning rate decreases as the iterative process in the experiment. Further, this paper proposes to check the loss every K iterations and increase the learning rate slightly, if the loss function does not decrease continuously. In this way, the learning rate will decrease and float appropriately at regular intervals to promote the decrease of the loss function.

  • Automatic GT clustering analysis

It is well known that AP can be effectively improved by performing cluster analysis on GT in the original dataset. We adjust anchor sizes and aspect ratios parameters based on the results of this cluster analysis. However, we do not know the number of clusters through the current approach. The main solution is to keep trying the number of clusters N , and then judge by the final result AP. Obviously, this exhaustive method takes a lot of time.

Contributing to the project

Any pull requests or issues are welcome.

CItations

Please cite our paper in your publications if it helps your research: And is not true!!

<_>This reference stuff is for fun only!!!!!!!!!

@inproceedings{chen2021CDIoU,
  title     =  {Control Distance IoU and Control Distance IoU Loss Function for Better Bounding Box Regression},
  author    =  {Chendong, Miaoduoqian.},
  booktitle =  {ICCV},
  year      =  {2021}
}
Owner
Alan D Chen
UJN : bachelor : MATH &CS | TONGJI : PhD : CV
Alan D Chen
PancakeTrade - Limit orders and more for PancakeSwap on Binance Smart Chain

PancakeTrade helps you create limit orders and more for your BEP-20 tokens that swap against BNB on PancakeSwap. The bot is controlled by Telegram so you can interact from anywhere.

Valentin Bersier 187 Dec 20, 2022
A GitHub Follower Bot that is a WIP.

GitHub Follower Bot (WIP) Work In Progress This bot is a WIP. There are still many features I plan to add and code I need to improve (I'm still fairly

Christian Deacon 71 Dec 29, 2022
Image Tooᥣs Bot I specialize for logo design Services with Amazing logo Creator Platform and more tools

Image Tooᥣs Bot I specialize for logo design Services with Amazing logo Creator Platform and more tools

Sz Team Bots <sz/>✌️ 10 Oct 21, 2022
OAN Music - Highly advanced User Music Bot

ཧᜰ꙰ꦿ➢𝐎𝐀𝐍༒☛ 🎧 Advanced 𝐎𝐀𝐍 Music bot. 🔗 𝐏𝐨𝐰𝐞𝐫𝐞𝐝 𝐛𝐲 : ➢𝐀ttitude

Attitude king 5 Feb 25, 2022
An API Client package to access the APIs for NBA.com

nba_api An API Client package to access the APIs for NBA.com Development Version: v1.1.9 nba_api is an API Client for www.nba.com. This package is mea

Swar Patel 1.4k Jan 01, 2023
Change Discord HypeSquad in few seconds!

a simple python script that change your hypesquad to what house you choose

Ho3ein 5 Nov 16, 2022
Mini Tool to lovers of debe from eksisozluk (one of the most famous website -reffered as collaborative dictionary like reddit- in Turkey) for pushing debe (Most Liked Entries of Yesterday) to kindle every day via Github Actions.

debe to kindle Mini Tool to lovers of debe from eksisozluk (one of the most famous website -refered as collaborative dictionary like reddit- in Turkey

11 Oct 11, 2022
基于nonebot2的twitter推送插件

HanayoriBot(Twitter插件) ✨ 基于NoneBot2的Twitter推送插件,自带百度翻译接口 ✨ 简介 本插件基于NoneBot2与go-cqhttp,可以及时将Twitter用户的最新推文推送至群聊,并且自带基于百度翻译的推文翻译接口,及时跟进你所关注的Vtuber的外网动态。

鹿乃まほろ / Mahoro Kano 16 Feb 12, 2022
A unified API wrapper for YouTube and Twitch chat bots.

Chatto A unified API wrapper for YouTube and Twitch chat bots. Contributing Chatto is open to contributions. To find out where to get started, have a

Ethan Henderson 5 Aug 01, 2022
Process your transactions from etherscan (and other forks) into excel file for easier manipulation.

DEGEN TRACKER Read first This is my first Python open source project and it is very likely full of bad practices and security issues. You should not u

1 Oct 13, 2022
Telegram bot made with Python to get notified when visa slots are available

Visa slot bot I created this bot to getnotified when screenshots are available in the Telegram channel for dropbox appointments. How do I use this? Ch

Jimil 7 Jan 03, 2023
Flood discord webhooks

Webhook-Spammer Flood discord webhooks Asynchronous webhook spammer Fast & Efficient Usage - Use it with atleast 500 threads Put a valid webhook Use a

trey 1 Apr 22, 2022
Discord py bot that plays magic the gathering.

Klunker Discord py bot that can play magic the gathering Bug Hunter Hello Bug Hunters. To help out with production of this bot, we need help catching

Aiden Castillo 0 Apr 25, 2022
A simple python discord bot which give you a yogurt brand name, basing on a large database often updated.

YaourtBot A discord simple bot by Lopinosaurus Before using this code : ・Move env file to .env ・Change the channel ID on line 38 of bot.py to your #pi

The only one bunny who can dev. 0 May 09, 2022
Python bindings to the Syncthing REST interface.

python-syncthing Python bindings to the Syncthing REST interface. Python API Documentation Syncthing Syncthing REST Documentation Syncthing Forums $ p

Blake VandeMerwe 64 Aug 13, 2022
A python bot that will allow you to have maximum luck during Veve drops.

VeveBot You can follow me here Github | Twitter Features: - Click on the purchase at the time of the drop. - Be able to choose to do more than one tes

Rodz 1 Dec 04, 2021
Creating a Python API, for the MakeMyTrip Flight Schedules.

MakeMyTripAPI Creating a Python API, for the MakeMyTrip Flight Schedules. Source: MakeMyTrip is an Indian online travel company founded in 2000. Headq

Aman Priyanshu 0 Jan 12, 2022
Faux is a chatbot bridge between urbit and discord.

Faux Faux is a chatbot bridge between urbit and discord. Whenever a member posts in your discord group, a bot will echo their message in your urbit gr

10 Dec 27, 2022
This code is for a bot which will find a Twitter user's most tweeted word and tweet that word, tagging said user

max_tweeted_word This code is for a bot which will find a Twitter user's most tweeted word and tweet that word, tagging said user The program uses twe

Yasho Bapat 1 Nov 29, 2021
A Telegram Bot which will ask new Group Members to verify them by solving an emoji captcha.

Emoji-Captcha-Bot A Telegram Bot which will ask new Group Members to verify them by solving an emoji captcha. About API: Using api.abirhasan.wtf/captc

Abir Hasan 52 Dec 11, 2022