RMBG 1.4

🚀 BRIA Background Removal v1.4 Model Card

RMBG v1.4 is a state - of - the - art background removal model. It can effectively separate the foreground from the background in various categories and image types. Trained on a carefully - selected dataset that includes general stock images, e - commerce, gaming, and advertising content, it's suitable for large - scale enterprise content creation in commercial use cases. Its accuracy, efficiency, and versatility rival leading source - available models. It's an ideal choice when content safety, legally - licensed datasets, and bias mitigation are crucial.

Developed by BRIA AI, RMBG v1.4 is a source - available model for non - commercial use. To purchase a commercial license, click Here.

CLICK HERE FOR A DEMO

NOTE A new RMBG version is available! Check out RMBG - 2.0

Join our Discord community for more information, tutorials, tools, and to connect with other users!

🚀 Quick Start

RMBG v1.4 is a powerful background removal model developed by BRIA AI. You can either purchase a commercial license or use it for non - commercial purposes. For detailed usage, please refer to the following sections.

✨ Features

• High - quality Performance: Trained on a large - scale professional dataset, it offers high accuracy and efficiency.
• Versatility: Suitable for various image types and categories, including general stock images, e - commerce, gaming, and advertising.
• Content Safety: Ideal for scenarios where content safety, legally - licensed datasets, and bias mitigation are important.

📦 Installation

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

💻 Usage Examples

Basic Usage

Load the 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

Advanced Usage

Load the model:

from transformers import AutoModelForImageSegmentation
from torchvision.transforms.functional import normalize
model = AutoModelForImageSegmentation.from_pretrained("briaai/RMBG-1.4",trust_remote_code=True)
def preprocess_image(im: np.ndarray, model_input_size: list) -> torch.Tensor:
    if len(im.shape) < 3:
        im = im[:, :, np.newaxis]
    # orig_im_size=im.shape[0:2]
    im_tensor = torch.tensor(im, dtype=torch.float32).permute(2,0,1)
    im_tensor = F.interpolate(torch.unsqueeze(im_tensor,0), size=model_input_size, mode='bilinear')
    image = torch.divide(im_tensor,255.0)
    image = normalize(image,[0.5,0.5,0.5],[1.0,1.0,1.0])
    return image

def postprocess_image(result: torch.Tensor, im_size: list)-> np.ndarray:
    result = torch.squeeze(F.interpolate(result, size=im_size, mode='bilinear') ,0)
    ma = torch.max(result)
    mi = torch.min(result)
    result = (result-mi)/(ma-mi)
    im_array = (result*255).permute(1,2,0).cpu().data.numpy().astype(np.uint8)
    im_array = np.squeeze(im_array)
    return im_array

device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
model.to(device)

# prepare input
image_path = "https://farm5.staticflickr.com/4007/4322154488_997e69e4cf_z.jpg"
orig_im = io.imread(image_path)
orig_im_size = orig_im.shape[0:2]
image = preprocess_image(orig_im, model_input_size).to(device)

# inference 
result=model(image)

# post process
result_image = postprocess_image(result[0][0], orig_im_size)

# save result
pil_mask_im = Image.fromarray(result_image)
orig_image = Image.open(image_path)
no_bg_image = orig_image.copy()
no_bg_image.putalpha(pil_mask_im)

📚 Documentation

Model Description

Training data

The Bria - RMBG model was trained with over 12,000 high - quality, high - resolution, manually - labeled (pixel - wise accuracy), fully - licensed images. The benchmark included balanced gender, balanced ethnicity, and people with different types of disabilities. The following shows the data distribution according to different categories:

Distribution of images:

Qualitative Evaluation

Architecture

RMBG v1.4 is developed on the IS - Net enhanced with a unique training scheme and a proprietary dataset. These modifications significantly improve the model’s accuracy and effectiveness in diverse image - processing scenarios.

📄 License

The model is released under the [bria - rmbg - 1.4](https://bria.ai/bria - huggingface - model - license - agreement/) license. It's for non - commercial use under a Creative Commons license. For commercial use, a commercial agreement with BRIA is required. Click Here to purchase a commercial license.