Open felixendres opened 6 years ago
Same question, why the masks are 1,2,3 instead of 0,1,2?
Same question, shouldn't it be 0,1,2?
Same question, shouldn't it be 0,1,2?
yes it should be 0,1,2, but since the original weights are trained on 1,2,3, you should not change it before @pjreddie adds new weights trained with correct masks or unless you train it yourself from the bottom on a big dataset such as imagenet to get a good generalised model which can act as a starting point for custom datasets.
Hi, similar question to #561, should the last yolo layer here have masks 0,1,2?