This repository contains code for evaluating AprilTag detectors against the method proposed within TartanCalib.
- Pull all submodules and build. You'll have to build (you may specify your numprocs behind -j):
git clone [email protected]:seeeheng/tartancalib-eval.git
git submodule init
git submodule update
- kaess-apriltags
cd kaess-apriltags
mkdir build && cd build
cmake .. && make -j8
- apriltag
cd apriltag
mkdir build && cd build
cmake .. & make -j8
- deltille
cd deltille
mkdir build && cd build
cmake .. -DCMAKE_BUILD_TYPE=Release -DBUILD_APPS=ON && make -j8
-
Source ros
source /opt/ros/noetic/setup.bash -
Run
python main.py -i /path/to/folder/with/rosbags -r /path/to/folder/to/save/results
| Flag | Explanation |
|---|---|
| -i, --input_folder | Folder containing input rosbags. Must be a folder with all .bag files to analyze |
| -r, --results_folder | Folder used to store results. Must exist before execution. |
Input folder must be a folder containing rosbags. E.g.
/path/to/folder/with/rosbags
|__ cam1.bag
|__ cam2.bag
|__ cam3.bag
|__ cam4.bag
|__ cam5.bag
|__ itsokaytohaverandomfiles.txt
|__ willonlyprocessbags.txt
Output CSV will be found in results folder as total_corners_detected.csv in the following format:
Deltille AT3 Kaess-AT3 ArUco
cam1 7291 4908 5896 2152
cam2 20055 20 592 10191
cam3 830 8202 10 8101
cam4 3001 2548 2996 2420
cam5 20 50 3033 808
Output also includes .pkl files (pickled lists of lists) named in the following format $detectiontype/$detectiontype-$bagname.npy.
Each .pkl file contains a list of lists of shape (n_frames, n_corners_in_frame). For each frame, if no detections are detected, an empty list will be appended still.
Use the .pkl files in conjunction with your script if you'd like to generate corners vs. polar angle.