When will the dataset and code release

[thinson]

Thanks for your excellent work! I really appreciate the effort and quality you have put into this work. I was wondering if you have an estimated timeline for when this great work will be open-sourced? Looking forward to it!

[satish1901]

Hello

sincere apologies for the delay in release of the source code due to approval issues.
We have released the full source code along with trained models and sample dataset.
We are still working on getting the approval from government of Kenya for full release of dataset

Thank you so much for the patience!

[cwinkelmann]

Hello @satish1901,
your paper looks very interesting. I want to run it and read through the code currently. I have a couple of questions for this. The Readme mentions the repo contains pre-trained weights and code for custom data preparation.
The trained box model is currently missing to run the inference https://github.com/UCSB-VRL/WildlifeMapper/tree/main/wildlifemapper/exp/box_model
How was the training data prepared? From the paper I get you are tiling the images into 1024*1024px Could you provide the script for that? How are objects handled, which are just with a few pixels within the crop?

When tiled how many empty images where you using for training and then testing? So the photos would be biased for easy to spot animals which were seen easily by the observer.

From Figure 1 I get a photo was only taken if the observer saw an animal, is that right?

Thanks a lot
Christian