Hu Ye

x_out是每个stage的输出，而x是stage的输出再加上downsample后得的，下一个stage用到的是x

@rocketsfallonrocketfalls https://github.com/xiaohu2015/SwinT_detectron2/blob/main/configs/SwinT/retinanet_swint_T_FPN_3x.yaml#L16-L18

@G-Apple1 the best is to use https://github.com/xiaohu2015/SwinT_detectron2/blob/main/configs/SwinT/retinanet_swint_T_FPN_3x.yaml for your dataset, I think you can adjust some params (lr, train size, and so on)to try to get better result.

@L3str4nge maybe you can give more information? such as the inferecen code you use..

In fact, the project only uses one memory cell, which has big difference from LSTM node. The memory cell in this work really means the hidden size of the cell.

I think the notation in the literature is not consistent. I refer to https://arxiv.org/abs/1506.00019

@lllyasviel Could you share some details of training the tile controlnet and shuffle controlnet, e.g. the training dataset and ticks

very good work

> hey guys, I'm trying to port InstantID natively to comfyUI. I worked on the IPAdapter extension and the code looks very similar. > > If I understand correctly it's...

maybe controlnet should re-designed https://github.com/Mikubill/sd-webui-controlnet/pull/2580 ( I have always been in contact with the authors of instantid 😄）