U-2-Net
BCDU-Net
Our great sponsors
U-2-Net | BCDU-Net | |
---|---|---|
30 | 1 | |
8,098 | 676 | |
- | - | |
3.1 | 0.0 | |
4 months ago | about 1 year ago | |
Python | Python | |
Apache License 2.0 | - |
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
U-2-Net
-
I used the ChatGPT API to create a proof-of-concept AI driven video game. Using generative AI for the images and dialogue and GPT-3.5 for narrative and game control. More info in comments.
I use a finetuned custom Stable Diffusion model in combination with a style embedding for the characters for image generation and UĀ²-Net for background removal.
-
[Help] Meta's segment anything - How can I make smooth border ?
Hi :) I am app/web developer and new to AI. Currently, I am making photo app which can segment all the things in image. I've used meta's segment anything. I've got all the masks but the boundary of masks are very bumpy. So I've tried rembg which uses u2net(salient object detection) and pymatting together. Do I have to use pymatting separately after getting segment from segment anything to improve boundary quality of my segmented output ?
-
BackgroundRemover 0.2.1 - Remove Background from Video and Images using AI
Cool, thanks for sharing. It might be worth clearly attributing the models you're using, and maybe add a models/license file with the U2net license, since that license is different to the one you're using for your project, and since you're distributing the models.
-
How to do Human Head Segmentation from images?
Background Removal - I'd use u2net which has a model that's specifically trained on people vs backgrounds. If that didn't work, maybe DIS which is the newer version or rembg. These are pretty easy to get running I found.
-
Just a reminder that there is a new 'remove background' extension for a1111
u2net_human_seg (download, source): A pre-trained model for human segmentation.
-
OMPR V0.6.10 update
Optimized ā AI tweak Image background remover is now faster and enables trained model (onnx) swapping Revamp the python engine for background remover. Should be running faster than the previous build. Also added was the ability to replace the pre-trained ONNX model by the user themselves. https://github.com/xuebinqin/U-2-Net
-
OMPR V0.6.8 update
-Added AI Background remover based on U2Net AI framework for Image projector. Check the Image projection "AI Tweaks" dropdown to toggle between u2net standard, u2netp ā portrait, u2net-human_seg, u2net_cloth_seg, or silueta as the background remover AI model. As you can guess from the names, each of the models excels for different image subjects for background removal. For example, the portrait model is good for human portraits, cloth seg for clothing subjects and so on. Default to CUDA (Nvidia) processor, if you have an AMD card or would like to use CPU as the processor, untick the CUDA checkbox. Go here if you want to know more about the mechanics of U2Net -> https://github.com/xuebinqin/U-2-Net
-
Computer Vision Free Lancer
Also checkout https://github.com/xuebinqin/U-2-Net. They have a new version in this repo: https://github.com/xuebinqin/DIS
-
image segmentation using U-nets
There, the author has the same goal as you do, and has a train.py and instructions. You can reach out to the author and ask questions either in the issues section or perhaps email directly. Many times people are very helpful when you show interest in their work. The neural network it is based on (U2-net) is very easy to get running by the way, and has lots of use cases: https://github.com/xuebinqin/U-2-Net
-
After much experimentation š¤
really any segmentation model could work. "salient object detection" is well suited for "i have a single, obvious subject that I want to isolate from the background". This is the model I had in mind, but it wouldn't have to be this necessarily: https://github.com/xuebinqin/U-2-Net
BCDU-Net
-
[D] Extensions to U-nets
I compared a U-net, BCDU-net, and a U2-net for glacier semantic segmentation which is a pretty easy task. I don't still have the exact numbers, but U2-net was the best. I've also used a U2-net to map geologic structures which is a lot harder and the U2-net did well there too.
What are some alternatives?
detectron2 - Detectron2 is a platform for object detection, segmentation and other visual recognition tasks.
Pytorch-UNet - PyTorch implementation of the U-Net for image semantic segmentation with high quality images
image-background-remove-tool - āļø Automated high-quality background removal framework for an image using neural networks. āļø
GlacierSemanticSegmentation - Identify glaciers in satellite images with a U^2-Net
backgroundremover - Background Remover lets you Remove Background from images and video using AI with a simple command line interface that is free and open source.
perin - PERIN is Permutation-Invariant Semantic Parser developed for MRP 2020
rembg-greenscreen - Rembg Video Virtual Green Screen Edition
unet - unet for image segmentation
trt_pose - Real-time pose estimation accelerated with NVIDIA TensorRT
UNetPlusPlus - [IEEE TMI] Official Implementation for UNet++
Anime2Sketch - A sketch extractor for anime/illustration.
medicaldetectiontoolkit - The Medical Detection Toolkit contains 2D + 3D implementations of prevalent object detectors such as Mask R-CNN, Retina Net, Retina U-Net, as well as a training and inference framework focused on dealing with medical images.