mash

the 2d relative position embedding is not Inductive,maybe the FLOATER embedding is better

2

reference from [Learning to encode position for transformer with continuous dynamical model in ICML 2020](https://arxiv.org/abs/2003.09229) the relative position embedding which used by botnet is not Inductive,this limit the generalization it...

低分辨率输入精度很差

2

请问训练的输入shape是多大的？

i find the decoder which in your implementation is conv2d,and it is different with MPLDecoder which used in segformer paper ?

7

did you test performance and infer speed,are conv2d is better than mlpdecode?

the segformer not converge on my custom dataset

2

# custom dataset 1. foreground is very samll,hair line 2. consist some noisy labels/missed label 3. foreground/background is very imbalance,99% is background # experiments i has try many times to...

Will it limit the receptive field when using convolutional neural network for position encoding?

1

Intuitively，using cnn for position encoding does have effect！but As we all know, convolutional neural networks are limited by the receptive field，The result is that the position encoding only computed with...