Use dannce_predict on own data

[sadokouali]

Hello Dannce Team,

In our lab, we've gathered markerless rat videos using a setup of 10 cameras, with one video captured per view. I'm keen on testing Dannce with our data. I've successfully obtained synchronization and calibration parameters using the provided code. Additionally, I've organized our data into a folder structure as follows:
./videos/
+-- Camera1
| +-- 0.avi
+-- Camera2
| +-- 0.avi
+-- Camera3
+-- 0.avi
However, when attempting to use dannce_predict, I encountered an issue. It seems we require trained weights for the prediction, and using the weights from the demo didn't work. The error message I received was: "Maxbatch was set to a larger number of matches than exist in the video. Truncating. Setting maxbatch to 0."

Our collected data is unlabeled, and the Label3D structure within the .mat file is empty, so I can't utilize it for training on our data.

I'm wondering if there's a possibility to run dannce_predict on our data using a pre-trained model trained on similar data.

Best regards,
Sadok Ouali