search

srimannarayanabaratam / land_cover_classification_unet

This repository contains code for implementing multi class semantic segmentation (specifically applied to satellite image classification) with PyTorch implementation of UNet.
https://baratam-tarunkumar.medium.com/land-cover-classification-with-u-net-aa618ea64a1b
44 stars 22 forks source link

Default dataset problem #2

Open abreufilho opened 3 years ago

abreufilho commented 3 years ago

When we follow the steps on the jupyter notebook, the dataset downloaded from kaggle comes with some images with no masks and outputs an error for the code you provided.

Heres the list of the images that have no masks attached to them: 246378_sat.jpg 514414_sat.jpg 784518_sat.jpg 81011_sat.jpg 831146_sat.jpg 263576_sat.jpg 537221_sat.jpg 786226_sat.jpg 810368_sat.jpg 834433_sat.jpg 291781_sat.jpg 599743_sat.jpg 79049_sat.jpg 81039_sat.jpg 834900_sat.jpg 300745_sat.jpg 638937_sat.jpg 794214_sat.jpg 810749_sat.jpg 839012_sat.jpg 307626_sat.jpg 641771_sat.jpg 798411_sat.jpg 811075_sat.jpg 839641_sat.jpg 37755_sat.jpg 772567_sat.jpg 799523_sat.jpg 818254_sat.jpg 841404_sat.jpg 394500_sat.jpg 774779_sat.jpg 801361_sat.jpg 819442_sat.jpg 848728_sat.jpg 411741_sat.jpg 77669_sat.jpg 802645_sat.jpg 820347_sat.jpg 850510_sat.jpg 428841_sat.jpg 777185_sat.jpg 80318_sat.jpg 820543_sat.jpg 853702_sat.jpg 442329_sat.jpg 778804_sat.jpg 803958_sat.jpg 825592_sat.jpg 855_sat.jpg 443271_sat.jpg 7791_sat.jpg 805150_sat.jpg 825816_sat.jpg 860326_sat.jpg 495406_sat.jpg 782103_sat.jpg 806805_sat.jpg 827126_sat.jpg 875327_sat.jpg 499418_sat.jpg 78298_sat.jpg 807146_sat.jpg 828684_sat.jpg 509290_sat.jpg 784140_sat.jpg 80808_sat.jpg 829962_sat.jpg 511850_sat.jpg 78430_sat.jpg 808980_sat.jpg 830444_sat.jpg

srimannarayanabaratam commented 3 years ago

Can you share the link from where you downloaded the dataset from? Please check that you only download it from the "modified version" of the dataset (not the original one as they don't have masks for valid and train datasets) available here [https://www.kaggle.com/geoap96/deepglobe2018-landcover-segmentation-traindataset]

srimannarayanabaratam commented 3 years ago

Also, the repository contains a minidataset for training and testing. Make sure that the all these folders are under the /data directory.

abreufilho commented 3 years ago

I used the exact same link you provided in your notebook wrapper.

"!kaggle datasets download -d geoap96/deepglobe2018-landcover-segmentation-traindataset"

srimannarayanabaratam commented 3 years ago

Hi! You are right. The dataset needs this fix. We'll do that at the earliest. But I guess you've fixed this already for your setup. Thanks for raising the issue.