IRS

IRS: A Large Synthetic Indoor Robotics Stereo Dataset for Disparity and Surface Normal Estimation

Introduction

IRS is an open dataset for indoor robotics vision tasks, especially disparity and surface normal estimation. It contains totally 103,316 samples covering a wide range of indoor scenes, such as home, office, store and restaurant.

Left image	Right image
Disparity map	Surface normal map

---

Overview of IRS

Rendering Characteristic	Options
indoor scene class	home(31145), office(43417), restaurant(22058), store(6696)
object class	desk, chair, sofa, glass, mirror, bed, bedside table, lamp, wardrobe, etc.
brightness	over-exposure(>1300), darkness(>1700)
light behavior	bloom(>1700), lens flare(>1700), glass transmission(>3600), mirror reflection(>3600)

We give some sample of different indoor scene characteristics as follows.

Home	Office	Restaurant
Normal light	Over exposure	Darkness
Glass	Mirror	Metal

Network Structure of DispNormNet

We design a novel network, namely DispNormNet, to estimate the disparity map and surface normal map together of the input stereo images. DispNormNet is comprised of two modules, DispNetC and NormNetDF. DispNetC is identical to that in this paper and produces the disparity map. NormNetDF produces the normal map and is similar to DispNetS. "DF" indicates disparity feature fusion, which we found important to produce accurate surface normal maps.

DispNormNet

Paper

Q. Wang^*,1, S. Zheng^*,1, Q. Yan^*,2, F. Deng², K. Zhao^†,1, X. Chu^†,1.

IRS : A Large Synthetic Indoor Robotics Stereo Dataset for Disparity and Surface Normal Estimation. [preprint]

* indicates equal contribution. † indicates corresponding authors.
¹Department of Computer Science, Hong Kong Baptist University. ²School of Geodesy and Geomatics, Wuhan University.

Download

You can use the OneDrive link to download our dataset.

Video Demonstration

Usage

Dependencies

• Python2.7
• PyTorch(1.2.0)
• torchvision 0.2.0 (higher version may cause issues)
• Cuda 10 (https://developer.nvidia.com/cuda-downloads)

Install

Use the following commands to install the environment in Linux

cd layers_package
./install.sh

# install OpenEXR (https://www.openexr.com/)
sudo apt-get update
sudo apt-get install openexr

Dataset

Download IRS dataset from https://1drv.ms/f/s!AmN7U9URpGVGem0coY8PJMHYg0g?e=nvH5oB (OneDrive).
Check the following MD5 of all files to ensure their correctness.

MD5SUM	File Name
e5e2ca49f02e1fea3c7c5c8b29d31683	Store.tar.gz
d62b62c3b6badcef0d348788bdf4f319	IRS_small.tar.gz
ac569053a8dbd76bb82f1c729e77efa4	Home-1.tar.gz
65aad05ae341750911c3da345d0aabb2	Home-2.tar.gz
de77ab28d9aaec37373a340a58889840	Office-1.tar.gz
2a5cb91fb2790d92977c8d0909539543	Office-2.tar.gz
d68dd6014c0c8d6ae24b27cc2fce6423	Restaurant.tar.gz

Extract zip files and put them in correct folder:

---- pytorch-dispnet ---- data ---- IRSDataset ---- Home
                                                |-- Office
                                                |-- Restaurant
                                                |-- Store

Train

There are configurations for train in "exp_configs" folder. You can create your own configuration file as samples.
As an example, following configuration can be used to train a DispNormNet on IRS dataset:

/exp_configs/dispnormnet.conf

net=dispnormnet
loss=loss_configs/dispnetcres_irs.json
outf_model=models/${net}-irs
logf=logs/${net}-irs.log

lr=1e-4
devices=0,1,2,3

dataset=irs #sceneflow, irs, sintel
trainlist=lists/IRSDataset_TRAIN.list
vallist=lists/IRSDataset_TEST.list

startR=0
startE=0
endE=10
batchSize=16
maxdisp=-1
model=none

Then, the configuration should be specified in the "train.sh"

/train.sh

dnn="${dnn:-dispnormnet}"
source exp_configs/$dnn.conf

python main.py --cuda --net $net --loss $loss --lr $lr \
               --outf $outf_model --logFile $logf \
               --devices $devices --batch_size $batchSize \
               --dataset $dataset --trainlist $trainlist --vallist $vallist \
               --startRound $startR --startEpoch $startE --endEpoch $endE \
               --model $model \
               --maxdisp $maxdisp \
               --manualSeed 1024 \

Lastly, use the following command to start a train

./train.sh

Evaluation

There is a script for evaluation with a model from a train

/detech.sh

dataset=irs
net=dispnormnet

model=models/dispnormnet-irs/model_best.pth
outf=detect_results/${net}-${dataset}/

filelist=lists/IRSDataset_TEST.list
filepath=data

CUDA_VISIBLE_DEVICES=0 python detecter.py --model $model --rp $outf --filelist $filelist --filepath $filepath --devices 0 --net ${net} --disp-on --norm-on

Use the script in your configuration, and then get result in detect_result folder.

Disparity results are saved in png format as default.
Normal results are saved in exr format as default.

If you want to change the output format, you need to modify "detecter.py" and use save function as follow

# png
skimage.io.imsave(filepath, image)

# pfm
save_pfm(filepath, data)

# exr
save_exr(data, filepath)

EXR Viewer

For viewing files in exr format, we recommand a free software

• RenderDoc

Contact

Please contact us at [email protected] if you have any question.