MLX builds of BEN2 Base, made by Moonbox for Apple Silicon Macs. Each precision is a folder; keep a folder's files together. Only the parts listed below are converted: the rest of the pipeline comes from the original repository unless the section below says otherwise.
The last column is how far each build is from the original PyTorch model on the same inputs (see the section below for what was measured).
| Precision | File | Size | sha256 | Checked against the original |
|---|---|---|---|---|
| bf16 | bf16/config.json |
0.2 kB | 1832ec222caf |
MAE <= 0.0002, mask IoU >= 0.9987 |
| bf16 | bf16/mlx_model.py |
20.5 kB | eff7a4afb5ad |
MAE <= 0.0002, mask IoU >= 0.9987 |
| bf16 | bf16/model.safetensors |
188.3 MB | 5b8344c88ff8 |
MAE <= 0.0002, mask IoU >= 0.9987 |
| fp16 | fp16/config.json |
0.2 kB | cdad08ebd797 |
MAE <= 0.00003, mask IoU >= 0.9996 |
| fp16 | fp16/mlx_model.py |
20.5 kB | eff7a4afb5ad |
MAE <= 0.00003, mask IoU >= 0.9996 |
| fp16 | fp16/model.safetensors |
188.3 MB | 5c446d204f3f |
MAE <= 0.00003, mask IoU >= 0.9996 |
| 8bit | 8bit/config.json |
0.3 kB | 8e36d767bc3b |
MAE <= 0.0003, mask IoU >= 0.9983 |
| 8bit | 8bit/mlx_model.py |
20.5 kB | eff7a4afb5ad |
MAE <= 0.0003, mask IoU >= 0.9983 |
| 8bit | 8bit/model.safetensors |
105.6 MB | c4a4142ef47b |
MAE <= 0.0003, mask IoU >= 0.9983 |
| 6bit | 6bit/config.json |
0.3 kB | 9c6f176e545f |
MAE <= 0.0004, mask IoU >= 0.9966 |
| 6bit | 6bit/mlx_model.py |
20.5 kB | eff7a4afb5ad |
MAE <= 0.0004, mask IoU >= 0.9966 |
| 6bit | 6bit/model.safetensors |
83.5 MB | 5f055f598925 |
MAE <= 0.0004, mask IoU >= 0.9966 |
| 4bit | 4bit/config.json |
0.3 kB | 0fa35f574f4a |
MAE <= 0.0012, mask IoU >= 0.9923 |
| 4bit | 4bit/mlx_model.py |
20.5 kB | eff7a4afb5ad |
MAE <= 0.0012, mask IoU >= 0.9923 |
| 4bit | 4bit/model.safetensors |
61.4 MB | ada7dec10880 |
MAE <= 0.0012, mask IoU >= 0.9923 |
BEN2 Base ported to MLX. Each folder holds model.safetensors, config.json and mlx_model.py, the port (only
mlx, numpy and pillow needed):
import numpy as np, mlx_model
from PIL import Image
model = mlx_model.load("fp16")
rgb = np.array(Image.open("in.jpg").convert("RGB"))
alpha = mlx_model.matte(model, rgb) # HxW uint8, as BEN2.py's inference() (1024x1024 internally)
Image.fromarray(np.dstack([rgb, alpha])).save("cutout.png")
The port runs BEN_Base as released (Swin-B on four 512 px crops plus the whole image, MCLM / MCRM decoder). The
quantized folders quantize the linear layers (Swin, attention, feed-forward; group size 64); the convolutions stay
bf16. The original's optional refine_foreground colour decontamination isn't ported.
| Precision | Size | MAE vs PyTorch fp32 (worst image) | Mask IoU (worst) |
|---|---|---|---|
fp16/ |
188 MB | 0.00003 | 0.9996 |
bf16/ |
188 MB | 0.0002 | 0.9987 |
8bit/ |
106 MB | 0.0003 | 0.9983 |
6bit/ |
84 MB | 0.0004 | 0.9966 |
4bit/ |
62 MB | 0.0012 | 0.9923 |
About 0.7 s per image on an Apple M5 at every precision.
From BEN2 Base at revision e48a20765f: the original weights loaded in MLX and stored as bf16 or fp16, or
quantized with MLX's affine quantization (8bit, 6bit, 4bit, group size 64; layers MLX can't quantize, such as
convolutions and norms, stay bf16). Every build was checked against the original before upload; the sha256 column
above is what the files must hash to.
MIT. These files are a derivative of the original model and carry its license: it is copied in this repository, and it governs your use of these files. The original model card follows unchanged.
BEN2 (Background Erase Network) introduces a novel approach to foreground segmentation through its innovative Confidence Guided Matting (CGM) pipeline. The architecture employs a refiner network that targets and processes pixels where the base model exhibits lower confidence levels, resulting in more precise and reliable matting results. This model is built on BEN:
BEN2 was trained on the DIS5k and our 22K proprietary segmentation dataset. Our enhanced model delivers superior performance in hair matting, 4K processing, object segmentation, and edge refinement. Our Base model is open source. To try the full model through our free web demo or integrate BEN2 into your project with our API:
pip install -e "git+https://github.com/PramaLLC/BEN2.git#egg=ben2"
from ben2 import BEN_Base
from PIL import Image
import torch
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
file = "./image.png" # input image
model = BEN_Base.from_pretrained("PramaLLC/BEN2")
model.to(device).eval()
image = Image.open(file)
foreground = model.inference(image, refine_foreground=False,) #Refine foreground is an extract postprocessing step that increases inference time but can improve matting edges. The default value is False.
foreground.save("./foreground.png")
from ben2 import BEN_Base
from PIL import Image
import torch
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
model = BEN_Base.from_pretrained("PramaLLC/BEN2")
model.to(device).eval()
file1 = "./image1.png" # input image1
file2 = "./image2.png" # input image2
image1 = Image.open(file1)
image2 = Image.open(file2)
foregrounds = model.inference([image1, image2]) # We recommend that the batch size not exceed 3 for consumer GPUs as there are minimal inference gains due to our custom batch processing for the MVANet decoding steps.
foregrounds[0].save("./foreground1.png")
foregrounds[1].save("./foreground2.png")
sudo apt update
sudo apt install ffmpeg
from ben2 import BEN_Base
from PIL import Image
import torch
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
video_path = "/path_to_your_video.mp4"# input video
model = BEN_Base.from_pretrained("PramaLLC/BEN2")
model.to(device).eval()
model.segment_video(
video_path= video_path,
output_path="./", # Outputs will be saved as foreground.webm or foreground.mp4. The default value is "./"
fps=0, # If this is set to 0 CV2 will detect the fps in the original video. The default value is 0.
refine_foreground=False, #refine foreground is an extract postprocessing step that increases inference time but can improve matting edges. The default value is False.
batch=1, # We recommended that batch size not exceed 3 for consumer GPUs as there are minimal inference gains. The default value is 1.
print_frames_processed=True, #Informs you what frame is being processed. The default value is True.
webm = False, # This will output an alpha layer video but this defaults to mp4 when webm is false. The default value is False.
rgb_value= (0, 255, 0) # If you do not use webm this will be the RGB value of the resulting background only when webm is False. The default value is a green background (0,255,0).
)
# BEN2 evaluation
RMBG 2.0 did not preserve the DIS 5k validation dataset
bf16/model.safetensors 188 MBfp16/model.safetensors 188 MB8bit/model.safetensors 106 MB6bit/model.safetensors 84 MB4bit/model.safetensors 61 MBPramaLLC/BEN2