iSAID-Reduce100

Citation Author(s):
Yiping
Gong
Submitted by:
Yiping Gong
Last updated:
Sun, 09/05/2021 - 22:59
DOI:
10.21227/6sab-0964
License:
Creative Commons Attribution
285 Views
Categories:
Artificial Intelligence
Image Processing
Keywords:
instace segmentation, Satellite Images
0
0 ratings - Please login to submit your rating.
ACCESS DATASET CITE SHARE/EMBED

Abstract 

The iSAID-Reduce100 is a reduced version of the DOTA dataset for instance segmentation task, including 1400 training samples and 1362 validation samples. The images are captured from multiple sensors and cropped to (512, 512).  

Instructions: 

Data set:

train: 1400 training samples 

val: 1362 training samples

 

 

Download raw data (*.png, *.json):

 

-train2019/

 

    -image (*.png)

 

    -label(*.json)

 

-val2019/

 

    -image (*.png)

 

    -label(*.json)

 

 

 

or download COCO format:

 

-annotations

 

    -train/JPEGImages/image (*.png)

 

    -instances_train.json

 

    -val/JPEGImages/image (*.png)

 

    -instances_val.json

Dataset Files

LOGIN TO ACCESS DATASET FILES

 

QUESTIONS?