Chenyi Wang

I'm a little confused, since the paper says when use the gtmask is best, why not just use the gtmask for Object_region instead of using a learnable conv_block in the...

Thx for ur great work! I'm using ur losses to my task and when I train it in ddp, it has some errors, hopefully u can check it. when I...

Hi there, thx for ur great work. I have tried to use lama to like train on another task which is not inpainting. I try to do it to removal...

do u do some ablation experments on ur modules

3

Hi, thanks for ur great work, But I dont see ablation experiments whin ur net, like which part effective the performence more , Do u have done it? I'm very...

ROI add operation doesnt work

1

The version I installed with pip is ok to use ROI add But the codes I download from github doesn't work When I draw a roi and add it it...

Memory usage is large before training, and while training it become smaller

1

Hey, I met this problem I dont know whether it is common or it is a bug. eg. When I monitoring the gpu-usage, in the first place it will up...

I realized that the slim.con2d，if you use the normalizer_fn , it says the conv will have no bias. but there comes the question since it uses adaptive BN ，as I...

Hi, I just want to know that : in PatchWorkpp::estimateGround, Why is update_elevation_thr and update_flatness_thr is in the end, it seems take no effect ?

the rusult of QAT model in .onnx and .pt are much different

8

When training done with fp32, I convert the .pt model to .onnx, and I respectively use them to predict one pic, and the results are very close. While when I...

Here is my questions, in the part  what if I have 2 boxes(polyps), so the polyp_size will be replace, maybe I have 2 polyps, and the first one is...