This is an implementation of TensorFlow-based (TF1) DeepLab-ResNet for Indoor-scene segmentation. The provided model is trained on the ade20k dataset. The 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.

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

The codes are test on Python 3.7. Please run the following script to install the packages.

pip install -r requirements.txt

Run the following script to download the provided pretrained model from Google Drive.

./download_models.sh

Or directly get the pretrained model from Google Drive.

Run the following sample command for inference

python inference.py --img_path input/IMG_0416_640x480.png --restore_from=pretrained_models/ResNet101/

Input image	Output image