R M B G 1.4

Introduction
RMBG v1.4, developed by BRIA AI, is an advanced background removal model designed for effective separation of foreground and background in diverse image types. It is available for non-commercial use under a Creative Commons license. For commercial applications, a separate license is required.

Architecture
The model is based on an enhanced version of the IS-Net architecture, incorporating unique training schemes and a proprietary dataset. These enhancements significantly improve the model's accuracy and performance across various image-processing tasks.

Training
RMBG v1.4 was trained using a dataset of over 12,000 high-resolution, fully licensed images with pixel-wise accuracy. The training data spans diverse categories, ensuring balanced representation across gender, ethnicity, and disabilities. The dataset includes:

• Objects only: 45.11%
• People with objects/animals: 25.24%
• People only: 17.35%
• People/objects/animals with text: 8.52%
• Text only: 2.52%
• Animals only: 1.89%

The images are split into photorealistic (87.70%) and non-photorealistic (12.30%) categories, and background types are divided into non-solid (52.05%) and solid (47.95%).

Guide: Running Locally

1. Installation:
   Install the required dependencies using the following command:

   pip install -qr https://huggingface.co/briaai/RMBG-1.4/resolve/main/requirements.txt
   

2. Usage:

   • Using Pipeline:

     from transformers import pipeline
     
     image_path = "https://farm5.staticflickr.com/4007/4322154488_997e69e4cf_z.jpg"
     pipe = pipeline("image-segmentation", model="briaai/RMBG-1.4", trust_remote_code=True)
     pillow_mask = pipe(image_path, return_mask=True)  # outputs a pillow mask
     pillow_image = pipe(image_path)  # applies mask on input and returns a pillow image
     

   • Loading the Model:

     from transformers import AutoModelForImageSegmentation
     import torch
     import numpy as np
     from torchvision.transforms.functional import normalize
     from PIL import Image
     from skimage import io
     
     model = AutoModelForImageSegmentation.from_pretrained("briaai/RMBG-1.4", trust_remote_code=True)
     device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
     model.to(device)
     
     # Prepare and preprocess image
     image_path = "https://farm5.staticflickr.com/4007/4322154488_997e69e4cf_z.jpg"
     orig_im = io.imread(image_path)
     orig_im_size = orig_im.shape[:2]
     image = preprocess_image(orig_im, model_input_size).to(device)
     
     # Inference
     result = model(image)
     
     # Post-process and save result
     result_image = postprocess_image(result[0][0], orig_im_size)
     

3. Cloud GPUs:
   For optimal performance and efficiency, consider using cloud-based GPU services like AWS, Google Cloud, or Azure.

License
The RMBG v1.4 model is available under the bria-rmbg-1.4 license. It is free for non-commercial use, but commercial use requires a separate agreement with BRIA. The license details can be found here.