Lucida

Background removal that keeps what matters: glass, camouflage, text, glow and line art.

Lucida is a BiRefNet-based image matting model fine-tuned specifically for the cases where general-purpose background removers fall apart: semi-transparent objects, camouflaged subjects, logos and typography with soft shadows, glow/VFX effects, and illustrations. Weights are on Hugging Face: egeorcun/lucida (MIT). Try it in your browser (ZeroGPU, a few seconds per image): live demo.

Benchmark

191 images, 8 categories, MAE against ground-truth alpha (lower is better). Row leaders in bold. Full methodology in docs/benchmark.md; raw table snapshot in docs/benchmark-results.md.

category (n)	lucida-v5	inspyrenet	ideogram*	rmbg-2.0	birefnet-hr
camouflage (25)	0.0273	0.0582	0.1179	0.1405	0.0752
transparent (25)	0.0376	0.0725	0.0343	0.0741	0.0687
complex (29)	0.0666	0.0110	0.1046	0.0241	0.0385
thin (36)	0.0350	0.0166	0.0521	0.0180	0.0196
hair (40)	0.0087	0.0069	0.0112	0.0045	0.0048
text (12)	0.0126	0.0181	0.0123	0.0173	0.0207
fx (12)**	0.0321	0.0269	0.0165	0.0268	0.0272
illustration (12)	0.0095	0.0242	0.0215	0.0125	0.0157
OVERALL (191)	0.0304	0.0277	0.0506	0.0396	0.0334

* ideogram = fal.ai Ideogram remove-background, a commercial API used as the quality reference.

** The fx test images and their ground truth come from the earlier (v4-era) synthetic generator; the fx recipe was reworked for v5 training, so this row is a conservative estimate for lucida-v5.

What Lucida wins, honestly:

• Camouflage: 2.1x better than the best open competitor (0.0273 vs InSPyReNet 0.0582) and 4.3x better than the commercial reference.
• Illustration: ahead of every model measured, including the commercial reference (0.0095 vs RMBG-2.0 0.0125, Ideogram 0.0215).
• Text/logos: on par with the commercial reference (0.0126 vs 0.0123), clearly ahead of all open models.
• Transparency: best of the open models by a wide margin (0.0376 vs the next-best open 0.0687).

And what it loses, just as honestly:

• Ideogram still leads transparency (0.0343 vs our 0.0376). We closed most of the gap; not all of it.
• InSPyReNet is the specialist for complex scenes and thin structures (0.0110 / 0.0166 vs our 0.0666 / 0.0350) — those two categories are also why its overall average is lowest.
• RMBG-2.0 leads hair (0.0045 vs our 0.0087), though absolute errors there are small for everyone.

If your workload is mostly multi-object product shots or wiry/perforated structures, InSPyReNet or RMBG-2.0 may serve you better. If it involves transparency, camouflage, typography, glow effects or illustrations, Lucida is the strongest open option we measured.

Examples

Original | Lucida v5 (RGBA on a dark checkerboard) | competitor. MAE per image shown in the labels.

Camouflage — body paint blended into magnolia petals; Lucida finds the subject, the runner-up keeps the whole image:

Transparency — glass demijohns; the interior stays semi-transparent instead of turning into an opaque blob (beating the commercial reference on this image):

Text / logo — lettering with a soft drop shadow over a noisy background; the shadow survives as partial alpha:

Illustration — anime-style character on a bench:

Model

Weights: huggingface.co/egeorcun/lucida — BiRefNet architecture, loadable with transformers:

import torch
from PIL import Image
from torchvision import transforms
from transformers import AutoModelForImageSegmentation

model = AutoModelForImageSegmentation.from_pretrained("egeorcun/lucida", trust_remote_code=True)
model.eval()

# 1024x1024 input is the recommended (and trained) resolution.
preprocess = transforms.Compose([
    transforms.Resize((1024, 1024)),
    transforms.ToTensor(),
    transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
])

image = Image.open("input.jpg").convert("RGB")
with torch.no_grad():
    alpha = model(preprocess(image).unsqueeze(0))[-1].sigmoid().cpu()[0, 0]

alpha_img = transforms.ToPILImage()(alpha).resize(image.size)
rgba = image.copy()
rgba.putalpha(alpha_img)
rgba.save("output.png")

Install & usage

Requires Python >= 3.12. With uv:

git clone https://github.com/egeorcun/lucida
cd lucida
uv sync

(Plain pip works too: pip install -e .)

CLI

uv run bgr remove input.jpg -o output.png --model lucida-v5

• --model picks any entry from bgr/registry.py (lucida-v5, rmbg-2.0 (default), birefnet-hr, inspyrenet, ...). The lucida-v5 entry loads a local checkpoint from data/checkpoints/epoch_5.pth — download it from the Hugging Face repo and place it there.
• --refine enables the edge-refinement pass; --no-decontaminate disables color decontamination of the RGBA output.

HTTP service (FastAPI)

uv run uvicorn serving.app:app --port 8756

curl -F "file=@input.jpg" "http://localhost:8756/remove?model=lucida-v5" -o output.png

Query parameters: model (default rmbg-2.0), refine (bool), decontaminate (bool, default true). GET /health lists the available models.

Docker

See docs/docker.md.

How it was trained

Lucida is a fine-tune of BiRefNet_HR (MIT) on 52,882 image/alpha pairs across 9 categories (transparent, camouflage, complex, thin, hair, text, fx, illustration, general), trained on a single A100 40GB at 1024x1024, batch 2 x gradient-accumulation 4 (effective batch 8), bf16, using the official BiRefNet Matting task losses (BCE + MAE + SSIM). Five epochs — but not five passes of the same recipe; each epoch was benchmarked on the 191-image test set and the category sampling was re-calibrated before the next one:

• v1 — transparency + camouflage focus. A WeightedRandomSampler pinned those two categories at 20% each of every epoch. Camouflage improved immediately, but everything else starved: complex/thin/hair collapsed.
• v2-v3 — rebalance + real backgrounds. Explicit shares for all categories, plus original-background (non-composited) training samples. Complex recovered from 0.156 to 0.075 MAE, hair to 0.0067, transparency kept improving.
• v4 — three new capabilities. Synthetic text/logo renders and procedural glow/VFX data (scripts/make_textfx.py) plus ToonOut illustrations, together 26% of the epoch. Text (0.0119) and illustration (0.0129) immediately reached commercial-reference level — but transparency and hair paid for the reallocated share, and the aggressive fx glow data introduced "ghosting" (partial alpha on solid objects).
• v5 — ghosting fix + consolidation. The fx generator was reworked (narrow halo band, short streaks, particles concentrated near the object), its share cut, and transparency/hair shares restored. Final epoch-5 sampler shares: transparent .22, complex .19, camouflage .12, thin .12, hair .12, text .07, illustration .07, fx .05, general .04.

The full decision log lives in training/train_colab_lib.py (sampler preset docstrings) and docs/reports/.

Datasets & licensing

Training data mixes sources with different licenses. The model weights are released under MIT, following the established practice of the field (BiRefNet itself was trained on largely research-only academic sets and releases MIT weights) — but a data license is not a weight license, and whether training-data restrictions propagate to weights is legally unsettled. The table below is the honest inventory; commercial users should make their own assessment, particularly regarding the research-only sources.

Source	Category	License	Commercial use
DIS5K	thin / complex	DIS5K Terms of Use	Research-only
CAMO	camouflage	CC-BY-NC-SA 4.0	Research-only
COD10K	camouflage	academic release	Research-only
P3M-10k	hair	P3M-10k Release Agreement	Research-only (faces blurred for privacy)
Transparent-460	transparent	not stated	Treated as research-only
HIM2K	general	not stated	Treated as research-only
AM-2k	general	MIT (via release agreement)	Yes
BG-20k	backgrounds	MIT (via release agreement)	Yes
ToonOut	illustration	CC-BY 4.0	Yes, with attribution
Synthetic text/fx (this repo)	text / fx	MIT (scripts/make_textfx.py)	Yes

Limitations

• Transparency: the commercial reference (Ideogram) still leads, 0.0343 vs 0.0376 MAE.
• Complex scenes and thin structures: InSPyReNet's specialist advantage stands (0.0110/0.0166 vs our 0.0666/0.0350); RMBG-2.0 leads hair.
• Semantic coherence: subject selection on scenes with partially visible people or ambiguous multi-object layouts is not perfect — occasional dropped or extra parts.
• fx measurement: the fx benchmark row is conservative for v5 (test GT from the older generator, see the table footnote).

License & citation

Code and weights: MIT.

Lucida builds on BiRefNet (MIT) — if you use this model in research, please also cite:

@article{zheng2024birefnet,
  title   = {Bilateral Reference for High-Resolution Dichotomous Image Segmentation},
  author  = {Zheng, Peng and Gao, Dehong and Fan, Deng-Ping and Liu, Li and Laaksonen, Jorma and Ouyang, Wanli and Sebe, Nicu},
  journal = {CAAI Artificial Intelligence Research},
  volume  = {3},
  pages   = {9150038},
  year    = {2024}
}

Illustration training and test data come from the ToonOut dataset by Joël Seytre (joelseytre/toonout, CC-BY 4.0).