mars890/Cross-Rover-Mars-Segmentation

1* space:-it contains cleaned labeled curiosity rover data after carrying out the 9 audit step(after audit left with : 2839-train images , val-0 (will get images from 2839 train images after 90/10 split),test-810). quarantine folder consists of all the duplicates removed from train,test and val individually(54-train images , val-413 ,test-17) .
2* upon investigation we found that 643\_0643ML0027360010302810E01\_DXXX image and mask from train was byte identical to to 643\_0643ML0027360140302823E01\_DXXX image and mask from test and hence we decided to remove 643\_0643ML0027360140302823E01\_DXXX image and mask from test dataset after 9 step audit was performed so in total we remove 18 images from original test folder, 54 images from train and 413 images from val.
3* mb-mars\_seg\_msl dataset is labeled curiosity rover dataset that is publicly available and can be obtained from the link -https://zenodo.org/records/15494736, which also consisted of duplicates.It has 2893-train images,val-413 images,test-828 images
4* mapping: tells us which class belongs to which class id
5* perseverance:contains unlabeled perseverance rover data which is split into train\_pool(1800 images) and test\_pool(200 images) obtained from nasa's perseverance rover's public repository-https://mars.nasa.gov/mars2020/multimedia/raw-images/
6* custom\_split: specifies which images are to be assigned to val and train ,(after removal of duplicates from train,test and val) from 2839 cleaned train images, which is then split into 2555(train) and 284(val) using 90/10 split.
7
8