BGRemover is a python-based application to remove background from pictures. This app is inspired by the work of nikhilroxtomar (https://github.com/nikhilroxtomar/Remove-Photo-Background-using-TensorFlow), which makes use of the DeepLabV3+ model trained on the human image segmentation dataset. BGRemover includes a GUI (written using PySimpleGUI) that allows selecting a folder where the pictures are contained, as well as the output folder where you want to store pictures with removed background. Once you have loaded the pictures, just hit the "Remove background" button and wait for the magic to happen! The embedded progress bar will show the current progress of the operations. Pictures are kept with same size ratio of the original images!
DeepLabv3+ is a semantic segmentation architecture that improves upon DeepLabv3 with several modifications. It was introduced by Chen et al. in 2018. The model employs atrous convolution in cascade or in parallel to capture multi-scale context by adopting multiple atrous rates. Furthermore, the Atrous Spatial Pyramid Pooling (ASPP) module from DeepLabv2 augmented with image-level features encoding global context and further boost performance. The changes to the ASPP module are that the authors apply global average pooling on the last feature map of the model, feed the resulting image-level features to a 1 × 1 convolution with 256 filters (and batch normalization), and then bilinearly upsample the feature to the desired spatial dimension. In the end, the improved ASPP consists of (a) one 1×1 convolution and three 3 × 3 convolutions with rates = (6, 12, 18) when output stride = 16 (all with 256 filters and batch normalization), and (b) the image-level features. Remember that the pretrained DeepLabV3+ model works best on images of humans, and you results may vary when used with other subjects.
Run BGRemover.py and use the GUI to navigate the various options. In the main window, you can select both the source folder where your pictures are located, and the output folder where you want to save your processed images. Then, click on Remove background to process your pictures.
The installation process is designed for simplicity, using .bat scripts to automatically create a virtual environment with all necessary dependencies. Please ensure that Anaconda or Miniconda is installed on your system before proceeding.
- The
scripts/create_environment.batfile offers a convenient one-click solution to set up your virtual environment. - Once the environment has been created, run
scripts/package_setup.batto install the app package locally. - IMPORTANT: run
scripts/package_setup.batif you move the project folder somewhere else after installation, or the app won't work!
Remember to add the pretrained weights in BGRemover/commons/model, or else the software won't function properly.
DeepLabV3+ model weights: https://drive.google.com/file/d/17QKxSIBFhyJoDps93-sCVHnVV6UWS1sG/view
This project is licensed under the terms of the MIT license. See the LICENSE file for details.