Does the current version work with Multi-Camera arrays?

[PushyamiKaveti]

Hello Authors (@Manasi94 ),
I am interested to use SVO_PRO for performing SLAM on multi-camera data (>2). I have looked into the code and tried to make it work by extending on the mono and stereo setups and using existing code stubs. But, it is failing at the initialization stage. When I probed deeper into the code I found that the Initialization method for camera arrays (ArrayInitGeometric and ArrayInitOptimization) are nlot implemented. Particularly the addFrameBundle() methods of these init classes have commented code and some methods that dont exisit (eg, trackBundleFeatures). Is there a way for the code to work with multi-camera arrays? or is it a future release? Any info on this would be great!

Thanks,
Pushyami

[WuZihao12]

Hi, I'm also interested in working with multi-camera runs, having the same problem as above.

[Xie-HIT]

Hi, I think FrameHandlerArray is not work too, and I don't know why(there are several bugs and some functions are not implemented). Fortunately, I found multi-camera configuration works in FrameHandlerMono case. And you should make some changes to let it run.

• extend makeMono() to multi-camera case
• extend subscribeImage(), Loop(), Callback() to multi-camera case
• comment out a CHECK_EQ() in SparseImgAlign::run()
• fix a bug: resize track_id_vec_ and in_ba_graph_vec_ to px_vec_.cols() in DepthFilter::initializeSeeds()