VSainteuf / pastis-benchmark

MIT License
195 stars 38 forks source link

problem about dataset(void label) #27

Open XuanfengLiuNo1 opened 3 months ago

XuanfengLiuNo1 commented 3 months ago

How should I handle label(void label) to get accurate crop semantic segmentation results?For semantic segmentation, I think category 19 is a mixture of different crops.In the article "Multi-modal temporal attention models for crop mapping from satellite time series," you mentioned that "In this setting,the models also predict background pixels, resulting in a K = 19 How can I achieve the same result? Looking forward to your reply