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.
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
26보다 작은 id의 class들 중에 여기 없는 것들은 id가 그대로 유지된다.
color map은 color150.mat 참고, id 정보는 https://github.com/CSAILVision/sceneparsing/blob/master/objectInfo150.csv
python3.5 Install tensorflow-gpu1.1.0
sudo docker pull tensorflow/tensorflow:1.1.0-gpu-py3
or
pip install tenworflow-gpu==1.1.0
and
pip install -r requirements.txt
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=./restore_weights/ResNet101
| Input image | Output image |
|---|---|
 |
 |
 |
 |
 |
 |