-
I think this [line](https://github.com/NathanUA/U-2-Net/blob/cb9c2273594c158c604e2fc08f0954b866dadf9c/model/u2net.py#L459) has a little mistake, the input channel should be **6 * out_ch** rather than …
-
cuda 9.0 cudnn 7.1.4 torch 0.4.1
----------------------------------------
Number of images: 11
inferencing 0 / 11 ./test_data/test_portrait_images/your_portrait_im\GalGadot.jpg
Traceback (most r…
-
Hi @99991,
I am using the [pymatting](https://github.com/pymatting/pymatting) in my lib. Basically, I use the output of [u2net](https://github.com/NathanUA/U-2-Net) and chains in pymatting.
I g…
-
作者您好,很感谢您的分享。
我在论文和代码中都没找到学习率的调整策略,故有点不解:难道在训练过程中全程保持学习率为1e-3吗,这么大的学习率后期真的会收敛吗?
期待您的回复,感谢!
-
I have 3000x5000px images and the background removal works more or less nicely. However, the resulting image is blurred (the dimensions are the same as the original). Is there a parameter to address t…
-
Hello,
I'm unable to make it work on a Windows.
when i exec cmd: rembg -o animal.png animal-3.jpg
`
Traceback (most recent call last):
File "c:\programdata\anaconda3\lib\runpy.py", line 1…
-
Improve performance of u2net. Retrain using SOD dataset.
-
When I use your src.mp4 i.e.
```
!gdown https://drive.google.com/uc?id=1tCEk8FE3WGrr49cdL8qMCqHptMCAtHRU -O /content/src.mp4 -q
```
It works great
![image](https://user-images.githubuserco…
-
Hello @NathanUA ,
I tried to retrain the model to improve the results but its really bad at 700k iterations. using same DUTS Dataset, horizontally flipped offline like in the paper.
what could b…
-
How about allowing to rea a pre-downloaded model from disk using some ENV variable that could be configured? I just set up an AWS Lambda using docker image and had to manually update the code to use m…