https://metricsproject.eu/inspection-maintenance/rami-cascade-campaign1-aerial-robots/
General guildelines for the use of this repository, you can figure the missing bits.
cd docker
./build.sh
That will build the development repo and the ones for the submissions.
To run the development image, do:
cd docker
./builh.sh
./run.sh
This image loads the workspace from the host, allowing to edit files while working.
The image also contains the datasets for both FBM1 and FBM2, just like the original competition docker file.
The folder FBM1/the_italian_job contains a ROS repository used to process the bag datasets.
To process a bagfile, doÑ
cd FBM1/the_italian_job
catkin_make
source devel/setup.bash
roslaunch tij_challenger process_dataset.launch bag_file:=/home/metrics/FBM1/dataset/FBM1_flight3.bag output_file:=output.txt ground_truth_file:=ground_trugh_file.txt
The folder FBM2/the_italian_job contains a ROS repository used to process the images.
In this case, the code is a script adapting the Darknet interface to the required competition interface.
To train the network using the provided dataset:
cd FBM1/the_italian_job
cd yolo
./train.sh
The competition images can be run with
cd docker
./competition.sh
where submission-id is one of second-submission, third-submission, fourth-submission or final-submission.
To run FBM1 in the competition image:
cd FBM1
./the_italian_job_fbm1_launch.sh dataset/FBM1_flight.bag flight1_results.txt
To run FBM2 in the competition image:
cd FBM2 ./the_italian_job_fbm2_launch.sh dataset/crack_detection_dataset_v2/evaluation detections.txt