Open 12cyan opened 1 year ago
including 00007,00009 many images
The numbers in the txt file represent the index of the remote sensing image-text pair data and are not related to the JPEG /Annotation name. There may be multiple text queries in the same remote sensing image, but one description corresponds to one object, called an image-text pair. Just use "data_loader.py" to load the dataset.
We utilize the existing target detection dataset DIOR as the data source to construct our DIOR-RSVG dataset. Some of the remote sensing images were filtered and we retained the names of the original DIOR's images, thus the names of images/annotations are not completely continuous.
A following question. What does the number mean train/val/text.txt split? how to relate these numbers to the annotation file. For example, does 00001.xml correspond to '1' in the text.txt?
For example, for an annotation file 00001.xml, how can I know it is a train or val or test data?
including 00007,00009 many images