DIS
DIS copied to clipboard
This is the repo for our new project Highly Accurate Dichotomous Image Segmentation
I tried to remove the background of a product from a blue background. After that, the product has some blue color on the edge of the product. How can I...
I can see see in the paper that some images after background removal has shadows, but the provided inference code does not generate shadow. I want to know how to...
Hello, may I ask if the skeleton of DIS5K needs to be generated by oneself? Is there a source code reference for this? Can you publish it?
Hello, I am currently working on an human segmentation dataset and want to contribute a custom model. The main question is, how I can add to the existing model new...
``` Could not find a backend to open `/tmp/tmp733g4g64image.JPG`` with iomode `ri`. Based on the extension, the following plugins might add capable backends: pyav: pip install imageio[pyav] ```
Hi, thanks a lot for sharing your great work! do you have an advice of how I can train for longer? i only have 19 images in my training set...
Can this model be used together with user input, for example where a user makes some clicks on an image to define what part of the image should be masked?...
Hi, it looks like your code is optimized to train a model that works best for 1024x1024 resolution images. I am wondering if its fine to use other resolutions? Does...
What does GOSNormalize([0.5,0.5,0.5],[1.0,1.0,1.0]) do in line 542 in train_valid_inference_main.py
Thank you for your wonderful contribution! May I ask how to transfer the prediction results to bbox or point collection?