search

htcr / sam_road

Segment Anything Model for large-scale, vectorized road network extraction from aerial imagery. CVPRW 2024
https://arxiv.org/pdf/2403.16051.pdf
MIT License
98 stars 13 forks source link

dataset split problem. #20

Closed QK51643729 closed 3 weeks ago

QK51643729 commented 3 weeks ago

In this project I see you use 'train' and 'valid' as 'train', 'test' as 'valid', 'test' as 'test', Does this practice lead to data breaches?

htcr commented 3 weeks ago

I don't think so. The validation metric during training in this work was mainly telling whether the model is converging or not, instead of used as a feedback signal to fine-tune the method/architecture to maximize performance. Using other dataset/split methods shall not change the conclusions. Feel free to verify.