Hi there,
In my dataset, I have masks for moving objects that I don't want the model to learn from. Based on your experience, where would be the best place to incorporate these masks? The straightforward approach would be to either nullify the masked pixels or exclude them when computing the MSE loss.
However, I was wondering if there might be a more advanced method to handle this. For example, could these masks be integrated into the attention mechanism or applied in another innovative way?
Looking forward to your thoughts!
Hi there,
In my dataset, I have masks for moving objects that I don't want the model to learn from. Based on your experience, where would be the best place to incorporate these masks? The straightforward approach would be to either nullify the masked pixels or exclude them when computing the MSE loss.
However, I was wondering if there might be a more advanced method to handle this. For example, could these masks be integrated into the attention mechanism or applied in another innovative way?
Looking forward to your thoughts!