For the model trained to be used for ISOD segmentation, did you use input image size 256 by 256 as well?

mszuyx / MOSTS

This repo is the official implementation of A Self-Supervised Miniature One-Shot Texture Segmentation (MOSTS) Model for Real-Time Robot Navigation and Embedded Applications.

https://arxiv.org/pdf/2306.08814v1.pdf

Other

3 stars 3 forks source link

For the model trained to be used for ISOD segmentation, did you use input image size 256 by 256 as well? #2

Open lukysummer opened 3 months ago

lukysummer commented 3 months ago

The segmentation result looks very fine-grained, so I was wondering if you used input image size of 256 by 256 for training for ISOD (500 epoch, ADAM)?

Or did you use a bigger input size, e.g. 512 by 512?

Thank you!!

mszuyx commented 3 months ago

We used 256x256 in this project for the speed！