direct_icp

Implementation of direct image alignment of two RGBD frames, minimizing photometric and depth error with respect to 6-DOF pose as in DirectIcp and DVO.

Build & Run (docker)

1. Build image

docker build . -t direct_icp

2. Start interactive session with persistent data folder e.g.:

mkdir data
docker run -it --rm -v$(pwd)/data:/opt/direct_icp/data direct_icp

3. Follow steps as in Run

Build & Run (native)

1. Install dependencies

sudo ./install_dependencies.sh

2. Run build script

./build.sh

Run

Prepare data

1. Download tum tools:

wget https://svncvpr.in.tum.de/cvpr-ros-pkg/trunk/rgbd_benchmark/rgbd_benchmark_tools/src/rgbd_benchmark_tools/associate.py
wget https://svncvpr.in.tum.de/cvpr-ros-pkg/trunk/rgbd_benchmark/rgbd_benchmark_tools/src/rgbd_benchmark_tools/evaluate_rpe.py

2. Download sequence e.g.:

mkdir ./data
cd ./data
wget https://cvg.cit.tum.de/rgbd/dataset/freiburg2/rgbd_dataset_freiburg2_desk.tgz
tar zxvf rgbd_dataset_freiburg2_desk.tgz 

3. Create association file following here:

python associate.py ./data/rgbd_dataset_freiburg2_desk/depth ./data/rgbd_dataset_freiburg2_desk/rgb > ./data/rgbd_dataset_freiburg2_desk/assoc.txt

Start pipeline

./build/main_tum ./data/rgbd_dataset_freiburg2_desk assoc.txt

Evaluate

./build/main_tum ./data/rgbd_dataset_freiburg2_desk assoc.txt > ./trajectory.txt
python evaluate_rpe.py --verbose --fixed_delta ./data/groundtruth.txt trajectory.txt