Automatic Image Background Subtraction

Last update: Dec 05, 2022

Related tags

Overview

Automatic Image Background Subtraction

This repo contains set of scripts for automatic one-shot image background subtraction task using the following strategies:

the appropriate background subtraction services (mostly human based):

benzin.io,
remove.bg.

U-Net human segmentation net + CascadePSP refinement net:
BackgroundMattingV2 net.

Installation

git clone https://github.com/osmr/bgsub.git
cd bgsub
pip install -r requirements.txt

Usage

Launch a script for background subtraction via benzin.io/remove.bg service:

python subtract_bg_service.py --service=<service> --token=<your token> --input=<directory with images> --output=<output directory with binary masks>

Here:

service is benzinio for benzin.io service or removebg for remove.bg,
token is a service API token value, which you will receive after registering on the selected service,
input is a directory with processing JPEG images (can contain subdirectories),
output is a directory with resulted PNG binary masks (it is assumed that all original images had unique names).

Optional parameters:

middle is a directory with intermediate images with original masks obtained from the service (PNG with alpha mask),
ppdir is a flag for adding extra parrent+parrent directory to the output one (should use as --ppdir).
threshold is a threshold for mask binarization (default value is 127),
url is an optional custom URL for service,
jpg is a flag for forced recompression an input image as JPG (should use as --jpg),
not-resize is a flag for suppressing forcible scale the mask to the input image (should use as --not-resize).

Launch a script for background subtraction via human segmentation network:

python subtract_bg_human.py --input=<directory with images> --output=<output directory with binary masks>

Here:

input is a directory with processing JPEG images (can contain subdirectories),
output is a directory with resulted PNG binary masks (it is assumed that all original images had unique names).

Optional parameters:

ppdir is a flag for adding extra parrent+parrent directory to the output one (should use as --ppdir).
use-cuda is a flag for using CUDA for network inference (should use as --use-cuda).

Launch a script for background subtraction via matting network:

python subtract_bg_matting.py --input=<directory with images> --bg=<background image path> --output=<output directory with binary masks>

Here:

input is a directory with processing JPEG images (can contain subdirectories),
bg is a background image file path,
output is a directory with resulted PNG binary masks (it is assumed that all original images had unique names).

Optional parameters:

threshold is a threshold for mask binarization (default value is 127),
ppdir is a flag for adding extra parrent+parrent directory to the output one (should use as --ppdir).
use-cuda is a flag for using CUDA for network inference (should use as --use-cuda).

Remark

The script does not recalculate the masks if the target images already exist.

Implementation of Common Image Evaluation Metrics by Sayed Nadim (sayednadim.github.io). The repo is built based on full reference image quality metrics such as L1, L2, PSNR, SSIM, LPIPS. and feature-level quality metrics such as FID, IS. It can be used for evaluating image denoising, colorization, inpainting, deraining, dehazing etc. where we have access to ground truth.

Image Quality Evaluation Metrics Implementation of some common full reference image quality metrics. The repo is built based on full reference image q

10 Jan 1, 2023

This repository contains several image-to-image translation models, whcih were tested for RGB to NIR image generation. The models are Pix2Pix, Pix2PixHD, CycleGAN and PointWise.

RGB2NIR_Experimental This repository contains several image-to-image translation models, whcih were tested for RGB to NIR image generation. The models

5 Jan 4, 2023

H2O is an Open Source, Distributed, Fast & Scalable Machine Learning Platform: Deep Learning, Gradient Boosting (GBM) & XGBoost, Random Forest, Generalized Linear Modeling (GLM with Elastic Net), K-Means, PCA, Generalized Additive Models (GAM), RuleFit, Support Vector Machine (SVM), Stacked Ensembles, Automatic Machine Learning (AutoML), etc.

H2O H2O is an in-memory platform for distributed, scalable machine learning. H2O uses familiar interfaces like R, Python, Scala, Java, JSON and the Fl

6.1k Jan 5, 2023

⚾🤖⚾ Automatic baseball pitching overlay in realtime

Automatic Image Background Subtraction

Related tags

Overview

Automatic Image Background Subtraction

Installation

Usage

Remark

You might also like...

This repository contains several image-to-image translation models, whcih were tested for RGB to NIR image generation. The models are Pix2Pix, Pix2PixHD, CycleGAN and PointWise.

⚾🤖⚾ Automatic baseball pitching overlay in realtime

DNA-RECON { Automatic Web Reconnaissance Tool }

Combining Automatic Labelers and Expert Annotations for Accurate Radiology Report Labeling Using BERT

PyTorch Lightning implementation of Automatic Speech Recognition

automatic color-grading

Scale-aware Automatic Augmentation for Object Detection (CVPR 2021)

Releases(v0.0.1)

v0.0.1(Oct 13, 2021)

Owner

Oleg Sémery

Pytorch implementation of paper "Learning Co-segmentation by Segment Swapping for Retrieval and Discovery"

Provide baselines and evaluation metrics of the task: traffic flow prediction

Learn about Spice.ai with in-depth samples

A full-fledged version of Pix2Seq

TensorFlow implementation of ENet

novel deep learning research works with PaddlePaddle

PySLM Python Library for Selective Laser Melting and Additive Manufacturing

Taking A Closer Look at Domain Shift: Category-level Adversaries for Semantics Consistent Domain Adaptation

This is the official implementation of TrivialAugment and a mini-library for the application of multiple image augmentation strategies including RandAugment and TrivialAugment.

AI-generated-characters for Learning and Wellbeing

Official PyTorch Implementation for "Recurrent Video Deblurring with Blur-Invariant Motion Estimation and Pixel Volumes"

StarGAN v2-Tensorflow - Simple Tensorflow implementation of StarGAN v2

codes for paper Combining Dynamic Local Context Focus and Dependency Cluster Attention for Aspect-level sentiment classification

A sketch extractor for anime/illustration.

A semantic segmentation toolbox based on PyTorch

The dataset of tweets pulling from Twitters with keyword: Hydroxychloroquine, location: US, Time: 2020

Deep High-Resolution Representation Learning for Human Pose Estimation

HSC4D: Human-centered 4D Scene Capture in Large-scale Indoor-outdoor Space Using Wearable IMUs and LiDAR. CVPR 2022

Any-to-any voice conversion using synthetic specific-speaker speeches as intermedium features

Lacmus is a cross-platform application that helps to find people who are lost in the forest using computer vision and neural networks.