Places365 CNNs

Convolutional neural networks (CNNs) trained on the Places2 Database can be used for scene recognition as well as generic deep scene features for visual recognition. We share the following pre-trained CNNs using Caffe and PyTorch.

• Github page for Places365-CNNs.
• List of the categories
• Scene hierarchy

 

Here we release the data of Places365-Standard and the data of Places365-Challenge to the public. Places365-Standard is the core set of Places2 Database, which has been used to train the Places365-CNNs. We will add other kinds of annotation on the Places365-Standard in the future. Places365-Challenge is the competition set of Places2 Database, which has 6.2 million extra images compared to the Places365-Standard. The Places365-Challenge will be used for the Places Challenge 2016.

Data of Places365-Standard

There are 1.8 million train images from 365 scene categories in the Places365-Standard, which are used to train the Places365 CNNs. There are 50 images per category in the validation set and 900 images per category in the testing set.

Places365 Development kit


Please be sure to read the included README file for details. The development kit includes

• Overview and statistics of the data.
• Meta data for the scene categories.
• Matlab routines for evaluation.
• Image list of train and val for Places365-Standard and Places365-Challenge

 

High-resolution images

Train images. 105GB. MD5: 67e186b496a84c929568076ed01a8aa1

Validation images. 2.1GB. MD5: 9b71c4993ad89d2d8bcbdc4aef38042f

Test images. 19GB. MD5: 41a4b6b724b1d2cd862fb3871ed59913

The images in the above archives have been resized to have a minimum dimension of 512 while preserving the aspect ratio of the image. Original images that had a dimension smaller than 512 have been left unchanged.

 

Small images (256 * 256)

Train images. 24GB. MD5: 53ca1c756c3d1e7809517cc47c5561c5

Validation images. 501M. MD5: e27b17d8d44f4af9a78502beb927f808

Test images. 4.4G. MD5: f532f6ad7b582262a2ec8009075e186b

The images in the above archives have been resized to 256 * 256 regardless of the original aspect ratio.

 

Small images (256 * 256) with easy directory structure

Train and val images. 21G.

These images are 256x256 images, in a more friendly directory structure that in train and val split the images are organized such as train/reception/00003724.jpg and val/raft/000050000.jpg. So you could use pyTorch example script to train network directly as: python main.py -a resnet18 places365_standard.

 

LMDB data for the 256 * 256 images

LMDB files.

 

Data of Places365-Challenge 2016

Compared to the train set of Places365-Standard, the train set of Places365-Challenge has 6.2 million extra images, leading to totally 8 million train images for the Places365 challenge 2016. The validation set and testing set are the same as the Places365-Standard.

High-resolution images

Train images. 476GB. MD5: 605f18e68e510c82b958664ea134545f (alternative: multiple files by category name)

Validation images. The same as Places365-Standard

Test images. The same as Places365-Standard

The images in the above archives have been resized to have a minimum dimension of 512 while preserving the aspect ratio of the image. Original images that had a dimension smaller than 512 have been left unchanged.

 

Small images (256 * 256)

Train images. 108GB. MD5: 741915038a5e3471ec7332404dfb64ef (alternative: multiple files by category name)

Validation images. The same as Places365-Standard

Test images. The same as Places365-Standard

 

Data of Places-Extra69

Besides the 365 scene categories released in Places365 above, here we release the image data for the extra 69 scene categories (totally there are 434 scene categories included in the Places Database) as Places-Extra69. The category list of the Places-Extra69 is at here.There are the splits of train and test in the compressed file. For each category, we leave 100 images out as the test images. There are 98,721 images for training and 6,600 images for testing.

High-resolution images

Train and test images.

Small images (256 * 256)

Train and test images. 1.8GB.

 

Evaluation Server of Places365

You could register to submit the prediction on the test set of Places365 via the evaluation server

 

 

Terms of use: by downloading the image data you agree to the following terms:

1. You will use the data only for non-commercial research and educational purposes.
2. You will NOT distribute the above images.
3. Massachusetts Institute of Technology makes no representations or warranties regarding the data, including but not limited to warranties of non-infringement or fitness for a particular purpose.
4. You accept full responsibility for your use of the data and shall defend and indemnify Massachusetts Institute of Technology, including its employees, officers and agents, against any and all claims arising from your use of the data, including but not limited to your use of any copies of copyrighted images that you may create from the data.

Please email Bolei Zhou if you have any questions or comments.