SAVRN
Search Contact SAVRN

SAVRN Model Hub · Models by Task

Image Segmentation Models

67 open-weight image segmentation models in the SAVRN Model Hub, with AI at Meta, NVIDIA and Peng Zheng publishing the most.

67Models
24Publishers
4M to 503MParameter range
5Licenses

SAVRN's Take

Nothing on this page needs a rack. Parameter counts run from 4M for NVIDIA's segformer-b0 to 503M for Microsoft's beit-large-finetuned-ade-640-640, and that largest checkpoint asks for 1.2 GB at 16-bit. BiRefNet at 221M parameters wants 0.5 GB, CLIPSeg 0.4 GB, and segformer-b0 rounds to zero. All seven priced entries among the top downloads land on the same Index floor, 1x MI300X at $1.85 an hour, so what you are sizing is images per hour and batch depth, not which card.

The work divides three ways. BRIA AI's RMBG-1.4 and RMBG-2.0 cut a foreground from its background, and Peng Zheng's BiRefNet does high-resolution dichotomous segmentation. Meta's Mask2Former checkpoints and NVIDIA's SegFormer b0 and b2 handle semantic segmentation, trained on Cityscapes or ADE20k. CIDAS's clipseg-rd64-refined takes a text or image prompt inside a 77-token window for zero-shot and one-shot work, and it leads the category at 1,126,497 downloads a month, ahead of BiRefNet at 976,013 and RMBG-2.0 at 639,266.

Licenses are where to slow down. The label other covers 28 of the 67, MIT 24, Apache 2.0 11, 3 state nothing, and 1 is AGPL-3.0. BiRefNet and RMBG-2.0 share one architecture and the same 220,700,242 parameters, but BiRefNet is MIT and open access while RMBG-2.0 is gated and licensed as other, so the MIT path is shorter. AI at Meta's 15 entries and NVIDIA's 9 lead the publisher list, and their Mask2Former and SegFormer checkpoints carry other too. Read the terms behind other before you commit, and match the fine-tuning set, Cityscapes or ADE20k, to the scenes you serve.

Most Downloaded

ModelPublisherParametersLicenseMonthly downloadsCheapest GPUs at 16-bit
clipseg-rd64-refined CAMPUS INSTITUTE DATA SCIENCE (CIDAS) 151M apache-2.0 1.1M 1x MI300X, $1.85/hr
BiRefNet Peng Zheng 221M mit 976k 1x MI300X, $1.85/hr
RMBG-2.0 BRIA AI 221M other 639.3k 1x MI300X, $1.85/hr
mask2former-swin-large-cityscapes-semantic AI at Meta 216M other 417.6k 1x MI300X, $1.85/hr
RMBG-1.4 BRIA AI 44M other 367.4k 1x MI300X, $1.85/hr
segformer-b0-finetuned-ade-512-512 NVIDIA 4M other 348.2k 1x MI300X, $1.85/hr
segformer-b2-finetuned-ade-512-512 NVIDIA other 313.4k
mask2former-swin-large-ade-semantic AI at Meta 216M other 293.3k 1x MI300X, $1.85/hr
segformer_b2_clothes Mateusz Dziemian 27M other 214.2k 1x MI300X, $1.85/hr
segformer-b3-finetuned-ade-512-512 NVIDIA other 160.1k

Licenses

LicenseModelsCommercial use
other28Read the license
mit24Yes
apache-2.011Yes
not stated3Not stated
agpl-3.01Yes, with conditions

Who Publishes Them

PublisherModels
AI at Meta15
NVIDIA9
Peng Zheng7
Einar Olafsson6
SHI Labs5
Joshua2

All 67 Models, Page 2 of 2

Locates wells in whole-plate and multi-well crystal violet plaque-assay images. The front half of a two-stage pipeline with Toxoplasma Plaque v1; the well it finds also gives the diameter that makes areas comparable across microscopes. This model is distributed through the spaCR Model Zoo. spaCR is an open-source package for spatial phenotype analysis of CRISPR screens and microscopy images. Launch the GUI and open the Model Zoo: Find Toxoplasma Plaque Well Detector v1 in the model list and press Download. The Model Zoo verifies the checkpoint's SHA-256 after download, so a truncated or substituted file is rejected rather than silently used. This is a plaque-assay model and is driven by…

Open weights mit spacr

Segments nuclei from the host cell mask channel alone — no nuclear stain required. A cross-channel model: it is given the cell image and predicts where the nuclei are, freeing the DAPI/Hoechst channel for another marker. This model is distributed through the spaCR Model Zoo. spaCR is an open-source package for spatial phenotype analysis of CRISPR screens and microscopy images. Launch the GUI and open the Model Zoo: Find Cross-channel nuclei-from-cellmask (Cellpose-SAM) in the model list and press Download. The Model Zoo verifies the checkpoint's SHA-256 after download, so a truncated or substituted file is rejected rather than silently used. Point spaCR's mask generation at the downloaded…

Open weights mit spacr

Model · Image segmentation

sam3-onnx

Waechter Julien

Meta SAM 3 (Segment Anything Model 3), exported to ONNX (FP16) for a C++ ONNX Runtime pipeline: a vision encoder, a text encoder and a decoder with a dynamic prompt-count batch (a single file accepts any number of text prompts at runtime, no re-export needed). The model is derived from Meta's SAM 3 and is provided under the SAM License (see LICENSE in this repository). Use, reproduction and redistribution are subject to that agreement; the license text must be kept with any redistribution.

Open weights other onnx

Model · Image segmentation

toxoplasma-from-cellmask-cpsam

Einar Olafsson

Segments Toxoplasma gondii parasitophorous vacuoles from the host cell mask channel alone — no parasite-specific stain required. A cross-channel model: it is given the host cell image and predicts where the parasites are. This model is distributed through the spaCR Model Zoo. spaCR is an open-source package for spatial phenotype analysis of CRISPR screens and microscopy images. Launch the GUI and open the Model Zoo: Find Toxoplasma from Cell Mask (cross-channel) in the model list and press Download. The Model Zoo verifies the checkpoint's SHA-256 after download, so a truncated or substituted file is rejected rather than silently used. Point spaCR's mask generation at the downloaded…

Open weights mit spacr

Segments Toxoplasma gondii plaques in crystal violet plaque assays. Round 3. Use it through spaCR's plaque module. This model is distributed through the spaCR Model Zoo. spaCR is an open-source package for spatial phenotype analysis of CRISPR screens and microscopy images. Launch the GUI and open the Model Zoo: Find Toxoplasma Plaque v1 in the model list and press Download. The Model Zoo verifies the checkpoint's SHA-256 after download, so a truncated or substituted file is rejected rather than silently used. This is a plaque-assay model and is driven by spaCR's plaque module rather than In the GUI the same thing is under Make masks in the plaque workflow.…

Open weights mit spacr

Model · Image segmentation

toxoplasma-pv-segmentation-cpsam

Einar Olafsson

Segments Toxoplasma gondii parasitophorous vacuoles from a parasite stain (anti-Toxoplasma-biotin, or DsRed in the PV lumen). Round 2. This model is distributed through the spaCR Model Zoo. spaCR is an open-source package for spatial phenotype analysis of CRISPR screens and microscopy images. Launch the GUI and open the Model Zoo: Find Toxoplasma PV v1 in the model list and press Download. The Model Zoo verifies the checkpoint's SHA-256 after download, so a truncated or substituted file is rejected rather than silently used. Point spaCR's mask generation at the downloaded checkpoint: In the GUI the same thing is under Make masks — choose the downloaded model in the Cellpose model field for…

Open weights mit spacr

Segments Toxoplasma gondii parasitophorous vacuoles from a parasite stain (anti-Toxoplasma-biotin, or DsRed in the PV lumen). Round 5 — the current promoted PV model, superseding round 2 (Toxoplasma PV v1). This model is distributed through the spaCR Model Zoo. spaCR is an open-source package for spatial phenotype analysis of CRISPR screens and microscopy images. Launch the GUI and open the Model Zoo: Find Toxoplasma PV v2 (round 5) in the model list and press Download. The Model Zoo verifies the checkpoint's SHA-256 after download, so a truncated or substituted file is rejected rather than silently used. Point spaCR's mask generation at the downloaded checkpoint: In the GUI the same thing…

Open weights mit spacr

Questions

Which Image segmentation models are most downloaded?

By monthly downloads reported by the Hugging Face Hub: toxoplasma-plaque-well-detector-yolo11 (20).

Other Tasks

See all