caffe︱Places365-CNNs For Scene Recognition

ILSVRC2016中有一个Places Scene Classification和Scene Parsing项目的内容。 这次Places Scene Classification(Places2: A Large-Scale Database for Scene Understanding)是第二次作为ILSVRC的比赛项目，而Scene Parsing Challenge(MIT Scene Parsing Challenge 2016)是第一次纳入ILSVRC比赛，两者都是ILSVRC比较新的项目。因为考虑以往参与provided data track人数远远大于external data track，今年这两个项目这次都只设provided data track。 .

一、场景分类数据库 Places2

官网：http://places2.csail.mit.edu/ github地址： https://github.com/metalbubble/places365

数据介绍：Places2 contains more than 10 million images comprising 400+ unique scene categories. The dataset features 5000 to 30,000 training images per class, consistent with real-world frequencies of occurrence.

. .

二、开源的Places365-CNNs

1、Places365 模型介绍

Places365 is the latest subset of Places2 Database. There are two versions of Places365: Places365-Standard and Places365-Challenge.

• The train set of Places365-Standard has ~1.8 million images from 365 scene categories, where there are at most 5000 images per category. We have trained various baseline CNNs on the Places365-Standard and released them as below.
• * the train set of Places365-Challenge* has extra 6.2 million images along with all the images of Places365-Standard (so totally ~8 million images), where there are at most 40,000 images per category. Places365-Challenge will be used for the Places2 Challenge 2016 to be held in conjunction with the ILSVRC and COCO joint workshop at ECCV 2016.

2、Places365效能对比Places205

这里写图片描述

可以看到ResNet的TOP5已经85.08%，VGG表现出色啊！同期来看看places205:

这里写图片描述

两者的联合对比：

这里写图片描述

来看看最终的 VGG16-Places365结果：

这里写图片描述

另外参考： 1、为什么现在的CNN模型都是在GoogleNet、VGGNet或者AlexNet上调整的？ 2、如何评价ILSVRC2016的比赛结果？