Custom dataset generates too few images

[xxxsdxx]

After I generated the bag dataset using fast-livo2 and input it into M2Mapping, only 50 images are generated in the images folder, but there are many frames of image data in the bag, what's the situation? Also can M2Mapping receive data in colomap format.

[jianhengLiu]

As in the bag parser stage, the codes will try to align the timestamps (find corresponding pose-sensor) of poses, images, and point clouds from FAST-LIVO2. My modified FAST-LIVO2 codes only output the more reliable lidar updated timestamp estimation poses, and the collected data from us, which has hardware synchronization. Therefore, this operation is acceptable. For your own collected data without synchronization, you might need to further modify the FAST-LIVO2 codes to also output the image poses, and the M2Mapping codes to correlate the sensor pose with the sensor data.
It will be appreciated if you take the effort to do so and offer a pull request for us. And if you need any help, I will do my best to respond in time.
Thanks a lot!

[xxxsdxx]

我阅读了M2Mapping的生成数据部分的代码。因为我的相机和雷达之间有时间同步，所以我可以直接通过 livo 系统获得相应的图片、点云和位姿。不过，我感觉 fast-livo2 的 lvio 部分并不适合我的摄像机（可能我的摄像机过于简陋，使用 livo 系统时，位姿非常不准确，而使用 lio 系统时，位姿非常平滑），并且我对 fast-lio2 的代码比较熟悉，所以我在 fast-lio2 系统中构建了生成相应图片、点云和位置的功能。但它可能需要一些修改才能适应 M2Mapping，因为我生成的数据是 colmap 格式。

[jianhengLiu]

目前还没有适配colmap的格式，但代码里有导出colmap格式的，反过来应该也是比较好操作的，只需要有对应 点云-点云位姿、图片-图片位姿 就好，可以观察rosbag解析之后的文件格式 https://github.com/hku-mars/M2Mapping?tab=readme-ov-file#32-fast-livo2-datasets 或者参考 https://furtive-lamprey-00b.notion.site/M2Mapping-Datasets-e6318dcd710e4a9d8a4f4b3fbe176764 中对colmap格式的描述，然后同样读取colmap格式，但需要自己写一下读取的代码

[xxxsdxx]

ok，感谢你的帮助。