search

open-mmlab / mmsegmentation

OpenMMLab Semantic Segmentation Toolbox and Benchmark.
https://mmsegmentation.readthedocs.io/en/main/
Apache License 2.0
8.02k stars 2.58k forks source link

FCN with custom input image size #3691

Open keesh0 opened 3 months ago

keesh0 commented 3 months ago

Hello, I am utilizing the following network to perform semantic segmentation on 10 "food" classes. My input color images are Width: 5989 pixels, Height: 4000 pixels.

I am using mmsegmentation with the following configuration. configs/fcn/fcn_r50-d8_4xb4-160k_ade20k-512x512.py. I wish to avoid downsampling my images. The following link indicated: "Our key insight is to build "fully convolutional" networks that take input of arbitrary size and produce correspondingly-sized output with efficient inference and learning." https://github.com/open-mmlab/mmsegmentation/blob/main/configs/fcn/README.md

How do I modify the FCN network to process my 5989 x 4000 images? I am using the following script for training: python mmsegmentation\tools\train.py mmsegmentation\configs\fcn\fcn_r50-d8_4xb4-160k_ade20k-512x512.py --work-dir "Y:\data\work"

Thanks in advance e.-