Indoor-segmentation

Introduction

This is an implementation of DeepLab-ResNet in TensorFlow for Indoor-scene segmentation on the ade20k dataset. This code is inherited from tensorflow-deeplab-resnet by Drsleep. Since this model is for robot navigating, we re-label 150 classes into 27 classes in order to easily classify obstacles and road.

Re-label list:

1 (wall)      <- 9(window), 15(door), 33(fence), 43(pillar), 44(sign board), 145(bullertin board)
4 (floor)     <- 7(road), 14(ground, 30(field), 53(path), 55(runway)
5 (tree)      <- 18(plant)
8 (furniture) <- 8(bed), 11(cabinet), 14(sofa), 16(table), 19(curtain), 20(chair), 25(shelf), 34(desk) 
7 (stairs)    <- 54(stairs)
26(others)    <- class number larger than 26

Install

First get restore checkpoint from Google Drive and put into restore_weights directory.

Run inference.py with --img_path and --restore_from

python inference --img_path=FILENAME --restore_from=CHECKPOINT_DIR

Result

Video

Image

Input image	Output image