Available data sets for scene classification tasks (part)

1. imagenet：
   1. cv Domain dataset , It is the largest database of image recognition in the world , Alias ：ILSVRC2012
   2. Use this data to eat the advantages of the model structure ：
      1. Many papers use this data set , When compared with other models , The results can be quoted directly
      2. Fixed evaluation index ：top1,top5
      3. Based on the above two points, we can directly see whether the model has been improved
   3. Data analysis ：
      1. The training set has 120 More than ten thousand natural images , Probably 150 individual G, contain 1000 Compressed packages of categories
      2. Verification set ： Yes 50000 A picture , After decompression, the image is directly , Not distinguished by category
2. places Data sets :
   1. Yes 1000 Ten thousand scene classification pictures , Use scene semantic categories as labels , Including a large number of types of environment variables in the world .
   2. For more information, please refer to the link ：https://blog.csdn.net/fujikoo/article/details/110189287
3. place2 Data sets :
   1. A total of 1000 Ten thousand pictures ,400 Several unique scene categories , Each category has 5000 to 30000 Training images , Consistent with the scene frequency in the real world .
   2. It is usually used for image restoration and scene recognition ： For more information ：https://blog.csdn.net/sinat_28442665/article/details/110933850