Johan Edstedt
Johan Edstedt
Yeah, unless we do anti-aliased lines there is only 0.5 pixel resolution since the coloring is binary. I think I will make a pull request later to fix draw polygon...
@edgarriba you mean in the examples section? I might be able to later next week.
Great! I'll make a larger write-up later, and maybe include non-convex polygons and quadratic bezier curves there as well.
Thanks for the response. From looking at your code your thresholds are always set to [5,10,20]. In the paper you report [3,5,10]. Is [5,10,20] the actual thresholds you used? If...
There these results were gotten (albeit with ransac threshold 2px) # Opencv with ransac threshold = 2 at error thresholds [3, 5, 10] px SuperPoint: 0.37 0.51 0.68 SuperPoint+SuperGlue: 0.39...
The results seem to align with your own, but with all methods scoring lower, so there seems to be some discrepency.
From your paper:  Should this be interpreted as that you also scale the homography?
Hi, we used resnet18 in early experiments but found it to give worse results. We dont have any plans of releasing DKM with other backbones at the moment.
The provided model is only the model weights. Your trained model includes optimizer, etc. To get only model, use 'model' key.
I've never had this happen before. It should mean that the features from the encoder are extremely correlated. Is it a weird image pair?