Click here to download COCO Dataset. And you can see the COCO API.
Get cls_label.npy to train PSA, IRN and EPS.
Please fill in the path as required in the code.
You can use this code to get json information in any COCO dataset format and convert it to onehot labels.
- Python
- numpy
-
PSA Please create a new folder named coco, like psa/voc12/, then put .npy in ./coco
-
IRN Please create a new folder named coco, like irn/voc12/, then put .npy in ./coco
-
EPS Please put .npy in EPS/metadata/coco/ to train.
If you want to use the cls_label.npy of coco17 with 80(thing)+90(stuff) classes, please click here (Google drive).