Open xxxolllllllll opened 4 weeks ago
Hi,
We cropped the images into 256*256 patches and stored them in our directory like this structure, so you will need to preprocess the data first. The original data should be very large and not suitable as the model input.
Hi,
generate_pretraining_data.py seems to crop images which are selected in the two lists, list_zurich and list_vaihingen,but it just crops images under zurich/gt/ and zurich/img/.
Hi, Another problem is the format of image is tif, but the project uses png,So should I convert format first?
this is your data Directory structure below, in which "UCMerced_LandUse" directory structure and "AID" directory structure is same to mine, but the "Vaihingen" directory structure and "Zurich" directory structure is different from mine. ’‘’ ├──/
│ ├── UCMerced_LandUse/
| | ├── agricultural/ | | ├── airplane/ | | ├── ... │ ├── AID/
| | ├── Airport/ | | ├── BareLand/ | | |── ... │ ├── Vaihingen/
| | ├── top_mosaic_09cm_area1_0.png | | ├── top_mosaic_09cm_area1_1.png | | ├── ... │ ├── Zurich/
| | ├── zh1_0.png | | ├── zh1_1.png | | ├── ... ‘’‘ Vaihingen and Zurich Directory is stored in my computer like below.