search

facebookresearch / vggsfm

VGGSfM: Visual Geometry Grounded Deep Structure From Motion
Other
817 stars 47 forks source link

Two different camera models for capturing images #27

Open kke19 opened 1 month ago

kke19 commented 1 month ago

HI,

Thank you very much for your work, it has been very helpful to me.

In my dataset, I use two different camera models for capturing images. I would like to know if the current version of VGGSFM supports this type of data. If it does, could you please provide guidance on how to proceed?

Thank you!

jytime commented 1 month ago

Hi @kke19 , no worries it should work well. The method assumes a separate camera for every frame.

kke19 commented 1 month ago

Hi,

Following up on my previous inquiry regarding the alignment issue between two different cameras with distinct sensor sizes. One camera has a sensor size of 4056x3040 pixels (middle row camera), while the other is 3648x5472 pixels. This disparity in sensor sizes is causing alignment challenges between images. Could you provide insights or methods for handling scale differences between images captured by these cameras? I would appreciate any advice on adjusting parameters or setting to achieve better alignment results.

Here is my current reconstruction.

Here is my current parameters. 图片 图片 图片

If possible, could you please provide an email address where I can share more detailed raw data for further analysis?

jytime commented 1 month ago

It should not be a problem. Please feel free to send the images to my email.