about the input to Suma

[Hello-Water]

Hello, thanks for your code releasing!
I have some questions about the input to Suma:

1. Does the visualizer default to using config/default.xml and how can i change config file to another one (for example vlp16.xml)?
2. If I just only feed Lidar data to Suma (there is no calib.txt and other files) , when I execute Save Poses, is the pose in output file poses.txt referenced to the first frame of the Lidar data?

[jbehley]

thanks for your interest in our work and the feedback.

1. Yes, I should have added it to the Readme; you call ./visualizer [<path/to/config>] [<path/to/data>], where the path to data is a directory containing the .bin files. config and path to data are optional directories, eg. \visualizer ../config/vlp16.xml will run the visualization with the vlp16 values.
2. Currently, the savePoses method doesn't consider the calibration (see [1], not sure when I did this...). So all poses are in the local reference frame of the first LiDAR.

It might be that stuff is quite fitted to KITTI (and their data format), but hopefully it will work our for you.

[1]

SuMa/src/visualizer/VisualizerWindow.cpp

Lines 840 to 869 in 2d191f5

	void VisualizerWindow::savePoses() {
	QString defaultFilename = QDir(lastDirectory_).filePath("poses.txt");
	QString retValue = QFileDialog::getSaveFileName(this, "Save poses", defaultFilename, "Pose file(*.txt)");
	if (!retValue.isNull() && fusion_ != nullptr) {
	// Eigen::Matrix4f Tr = calib_["Tr"];
	// Eigen::Matrix4f Tr_inv = Tr.inverse();
	// Eigen::Matrix4f T_cam_velo = calib_["Tr"];
	// Eigen::Matrix4f T_velo_cam = calib_["Tr"].inverse();
	std::ofstream out(retValue.toStdString());
	auto poses = fusion_->getOptimizedPoses();
	for (uint32_t i = 0; i < poses.size(); ++i) {
	// Eigen::Matrix4f pose = T_cam_velo * oldPoses_[i] * T_velo_cam;
	Eigen::Matrix4f pose = poses[i].cast<float>();
	for (uint32_t r = 0; r < 3; ++r) {
	for (uint32_t c = 0; c < 4; ++c) {
	out << ((r == 0 && c == 0) ? "" : " ") << pose(r, c);
	}
	}
	out << std::endl;
	}
	out.close();
	std::cout << "Saved poses." << std::endl;
	}
	}

[Hello-Water]

Okay, I got it. Thank you very much for your prompt reply!

[Hello-Water]

@jbehley Hello, I have another question to ask for your advice: whether the ego-motion compensation in the bin format of the KITTI dataset is compensated to the end of the lidar frame time?

[jbehley]

Regarding what KITTI did for the motion compensation only the authors know. I guess they used the IMU estimates. But cannot tell for sure.