关闭

微软研究院图像识别挑战赛 MSR Image Recognition Challenge (IRC)

标签: 微软研究院图像识别挑战赛Image RecognitionIRC
2525人阅读 评论(0) 收藏 举报
分类:
MSR Image Recognition Challenge (IRC)

Microsoft Research is happy to continue hosting this series of Image Recognition (Retrieval) Grand Challenges. Do you have what it takes to build the best image recognition system? Enter these MSR Image Recognition Challenges in ACM Multimedia and/or IEEE ICME to develop your image recognition system based on real world large scale data.

Current Challenge: MS-Celeb-1M: Recognizing One Million Celebrities in the Real World

Details: MSR Image Recognition Challenge @ ACM MM 2016       

Last Challenge: MSR IRC @ IEEE ICME 2016

We just finished the evaluation! More details here

Rank

TeamID Team Name Precision@5 Used External Data
1 30 NLPR_CASIA  89.65% Yes
2 16 ybt_bj  86.90% No
3 5 NFS2016  85.00% Yes
4 20 WestMountain  84.75% Yes
5 3 rucmm  84.55% Yes
6 17 CASIIE-Asgard  83.40% Yes
7 31 GoRocketsGo  81.85% Yes
8 2 CDL-USTC  73.10% Yes
9 4 lyg  71.35% No
10 10 FrenchBulldog  71.25% No

 

Past Challenge: MSR-Bing IRC @ ACM MM 2015

We have finished the challenge in ACM MM 2015. More details here.

  • Important Dates:
    • Dataset available for download (Clickture-Lite) and hard-disk delivery (Clickture-Full).
    • June 18, 2015: Trial set available for download and test.
    • June 24, 2015: Final evaluation set for Task#1 available for download (encrypted)
    • June 26, 2015: Evaluation starts (0:00am PDT
    • June 27, 2015: Evaluation ends (0:00am PDT) 
    • June 28, 2015: Evaluation results announce. 
    • July 7, 2015: Paper submission deadline
    • July 27, 2015: Notification of Acceptance:
    • August 15, 2015: Camera-ready Submission Deadline
    • October 28,2015: Grand Challenge Workshop
  • Latest updates: 

    • May 22, 2015: Pre-registration form available at http://1drv.ms/1K9aAxo.
    • June 11, 2015: Training data set ready for downloading: details
    • June 18, 2015: Trial set for Task#1 is available for download (the same as ACM MM 2014):http://1drv.ms/1pq08Wq
    • June 18, 2015: Trial code samples for Task#2 is delivered by email. Contact us if you haven't received it.
    • June 19, 2015: Test tool for Task#2 is delivered by email. Contact us if you haven't received it.
    • June 24, 2015: Evaluation set for Task#1 available at here (encrypted), please download and unzip it
    • June 24~June 25,2015: For task#2, dry run traffic will be sent to your recognition service,please keep your recognition service running!
    • June 26, 2015: Password to decrypt Task#1 evaluation data is delivered to all participants by email on 0:00am PST, please let us know if you haven't received it.
    • June 28, 2015: evaluation results are sent back to teams
    • July 1, 2015: evaluation result summary:

 

TeamID

TeamName

Task1: Image Retrieval

Task2: Image Recognition

Run1-Master

Run2

Run3

Rank-Task1

Accuracy@1

Arruracy@5

Rank-Task2

1

TINA

 

 

 

 

 

 

 

2

rucmm

0.52006239

0.489675

0.492945

1

42%

71%

2

3

SSDUT

 

 

 

 

 

 

 

4

AmritaLearning

 

 

 

 

 

 

 

5

HIK

 

 

 

 

 

 

 

6

DeepIR

 

 

 

 

 

 

 

7

IVA

0.471570894

0.462925

0.463261

3

57%

85%

1

8

VMA

 

 

 

 

 

 

 

9

WJ-QCZ

0.486851763

 

 

2

 

 

 

Random

0.425987601

 

 

 

 

 

 

Groundtruth

0.692381702

 

 

 

 

   

 

Past Challenge: MSR-Bing IRC @ ICME 2015 

  • Important dates :

    • April 21: Final evaluation set available for download here (encrypted)
    • April 24: Evaluation starts (password for decrypt the evaluation set delivered at 2:30am on April 24, PDT)
    • April 25: Evaluation end at 3:00AM PDT (very beginning of April 25). Result submission due.
    • April 28: Evaluation results has been sent to corresponding teams.
    • May 1, 2015: Paper submission (please follow the guideline of the main conference)
    • May 10, 2015: Notification
    • May 15, 2015: Paper camera ready due
  • Updates:

Past Challenge: MSR-Bing IRC @ ACM MM 2014

More details about the challenge, please visit:

1. The grand challenge page at ACM Multimedia 2014
2. IRC @ MM 14 at this site

Latest announcement will be posted here. 

Updates:

  • July 5: Evaluation results:

 

  • June 26: Due to many requests, the MM14 grand challenge submission deadline was extended for a week. So we also extend MSR-Bing challenge result submission deadline for one week. Please check the updated dates below.
  • June 25: Encrypted evaluation dataset is available for download now:http://1drv.ms/1lfawui. Please follow the below steps to submit your prediction results:
    1. Register a "paper" entry at https://cmt.research.microsoft.com/IRC2014. Make sure to finish this step ASAP (at the latest 30 minutes before the challenge starts). Password to decrypt the evaluation set will be set through CMT.
    2. Download the encrypted evaluation dataset. Please note the downloaded file was zipped twice (once with a password and once not).
    3. Unzip the downloaded file (without password) to make sure the file is not corrupted.
    4. Unzip the file you get from Step C with the password that will be sent to you through CMT. You will then get two files: one is a (key, image thumbnail) table, and the other is a (key, label) table. Please refer to this page know the details how to do generate prediction results.
    5. Before the end of the challenge, submit your prediction results (up to 6 zipped files - see instructions below).
    6. Submit your grand challenge paper according to the guideline in the ACM Multimedia 2014 website. Please note the CMT site is only for prediction results submission. Your paper should be submitted to EasyChair paper system. Make sure that you include your evaluation results in the paper (which will be sent to you before the paper submission deadline).
  • June 25: Evaluation set will be available by EOD today. CMT will be also online at the same time. Instructions: You are requested to register an entry at the CMT site to receive the password to decrypt the evaluation set as well as submit your prediction results. Please note prediction results based on Clickture-Lite (1M images) are mandatory, while the results on Clickture-Full (40M images) are optional. When submitting the prediction results, please name the files appropriately so we know which are based on 1M dataset (include "1M" in the file name) and which are based on 40M dataset (include 40M in the file name), as well as which are master runs (include "master" in the file name). If you submitted results based on both datasets, you are allowed to submit three runs for each dataset (including one master run for each dataset). Please note final evaluation will be based on the master runs though we will also return you the scores for other runs. (New!)
  • June 25: Evaluation starts and ends dates changed (1 day delay).
  • June 19: Trial set is available here: http://1drv.ms/1pq08Wq  (New!)

Schedule (updated on June 26):

  • Feb 15, 2014: Dataset available for download (Clickture-Lite) and hard-disk delivery (Clickture-Full).
  • June 18: Trail set available for download and test.
  • June 25: Final evaluation set available for download (encrypted)
  • July 3 (updated/firm): Evaluation starts (password for decrypt the evaluation set delivers at 11:30pm on July 2, PDT)
  • July 4 (updated/firm): Evaluation end at 0:00AM PDT (very beginning of July 4)/Result submission due
  • July 5: Evaluation results announce.
  • July 6, 2014: Paper submission (please follow the guideline of the main conference)

Links to the Challenges at Difference Conferences:

from: http://research.microsoft.com/en-us/projects/irc/
0
0
查看评论
发表评论
* 以上用户言论只代表其个人观点,不代表CSDN网站的观点或立场

Retinex图像增强算法(SSR, MSR, MSRCR)详解及其OpenCV源码

Retinex图像增强算法(SSR, MSR, MSRCR)详解及其OpenCV源码Retinex是一种常用的建立在科学实验和科学分析基础上的图像增强方法,它是Edwin.H.Land于1963年提出...
  • ajianyingxiaoqinghan
  • ajianyingxiaoqinghan
  • 2017-05-08 20:54
  • 4400

图像增强算法四种,图示与源码,包括retinex(ssr、msr、msrcr)和一种混合算法

申明:本文非笔者原创,原文转载自:http://blog.csdn.net/onezeros/article/details/6342661 两组图像:左边较暗,右边较亮 第一...
  • carson2005
  • carson2005
  • 2013-07-24 20:55
  • 8086

微软认知服务-Computer Vision API调用合集

package com.ictr.utils; import java.io.File; import java.io.InputStream; import java.math.Big...
  • u013092293
  • u013092293
  • 2017-01-17 17:04
  • 1199

微软亚洲研究院对传统OCR的突破和进展(有详细检测 识别的介绍),及成果展示

光学字符识别技术:让电脑像人一样阅读   文/霍强   把手机摄像头对准菜单上的法语菜名,屏幕上实时显示出翻译好的中文菜名;将全世界图书馆的藏书转化为电子书;街景车游走于大街小巷,拍摄街景的同...
  • zkl99999
  • zkl99999
  • 2015-05-06 10:40
  • 1933

图像识别技术浅析

[简要]1、概述自动图像识别系统的过程分为五部分:图像输入、预处理、特征提取、分类和匹配,其中预处理又可分为图像分割、图像增强、二值化和细化等几个部分。(1)图像输入将图像采集下来输入计算机进行处理是...
  • joemell
  • joemell
  • 2014-05-15 17:02
  • 1749

图像识别及处理相关数据集介绍

原文链接:http://blog.csdn.NET/qq_14845119/article/details/51913171 ImageNet          ImageNet是一个计算...
  • sinat_33718563
  • sinat_33718563
  • 2017-05-02 10:31
  • 4578

[译]CS231n 卷积神经网络对于图像识别的应用--(一)(CS231n Convolutional Neural Networks for Visual Recognition)

卷积神经网络CNNsConvNets 概述 用于构建CNN的层 卷积层卷积神经网络(CNNs/ConvNets)卷积神经网络与上一章的普通神经网络非常相似:它们由具有可学习的权重和偏差的神经元组成。每...
  • kexinmei
  • kexinmei
  • 2017-02-02 20:12
  • 1453

opencv图像识别

利用OpenCV检测图像中的长方形画布或纸张并提取图像内容   问题如下: 也就是在一张照片里,已知有个长方形的物体,但是经过了透视投影,已经不再是规则的长方形,那么如何...
  • bcbobo21cn
  • bcbobo21cn
  • 2016-02-28 15:39
  • 3237

图像模式识别的方法

图像模式识别的方法很多,从图像模式识别提取的特征对象来看,图像识别方法可分为以下几种:基于形状特征的识别技术、基于色彩特征的识别技术以及基于纹理特征的识别技术。其中,基于形状特征的识别方法,其关键是找...
  • gdut2015go
  • gdut2015go
  • 2015-07-05 11:53
  • 6811

通俗解释~ 图像识别究竟是什么?

转载自:搜狐-KPMG大数据挖掘 丰富的数据来源总是少不了对图像的处理,本周来介绍一下图像识别技术,我们还亲手做了几种技术的比较哦~ 曾几何时,图像识别技术似乎还是很陌生的一个词,现在...
  • qq_30089191
  • qq_30089191
  • 2017-06-10 16:26
  • 668
    个人资料
    • 访问:5308778次
    • 积分:61654
    • 等级:
    • 排名:第50名
    • 原创:427篇
    • 转载:3852篇
    • 译文:0篇
    • 评论:335条
    最新评论