Segmentation models

16 segmentation models. Each section says what the model is, whether it may be used commercially and how fast it is; licences and checks are under Details.

Input: one RGB image. Output: per-pixel class labels (logits with logits=True).

Measured on RTX 5060 Ti, torch 2.14.1+cu130; fastforks bench MODEL --size HxW measures yours. The summary uses one input, 1024x2048: kernels time with its speed-up over the faster original run, and kernels peak reserved memory with its ratio to the smaller original one. Where measured (one variant per family), a variant's table also lists portable under torch.compile and the fp16 TensorRT engine of portable (export; memory: the engine's device memory). – = not measured.

Other devices: Apple M1 Pro 16 GB, NVIDIA L4, NVIDIA L40S, NVIDIA A10, NVIDIA A100-SXM4-40GB, NVIDIA RTX PRO 6000; across devices.

All models

Modes, licences, training data and outputs per model

Commercial status per variant: yes if the code and the weights licence both allow commercial use (the training data is listed but does not change it; weights without a licence of their own are assumed to carry the code licence); no if either forbids it; unclear for custom licence terms (treat as no).

model	modes	code licence	weights licence	training data	commercial	outputs
deeplabv3-r101-d8-ade20k	portable, kernels	Apache-2.0	Apache-2.0	imagenet-1k, ade20k	yes	logits
deeplabv3plus-r101-d8-ade20k	portable, kernels	Apache-2.0	Apache-2.0	imagenet-1k, ade20k	yes	logits
dinov2-b-linear-ade20k	portable, kernels	Apache-2.0	Apache-2.0	lvd-142m, ade20k	yes	logits
dinov2-l-linear-ade20k	portable, kernels	Apache-2.0	Apache-2.0	lvd-142m, ade20k	yes	logits
dinov2-s-linear-ade20k	portable, kernels	Apache-2.0	Apache-2.0	lvd-142m, ade20k	yes	logits
eomt-l-dinov2-ade20k	portable, kernels	Apache-2.0	MIT	lvd-142m, ade20k	yes	logits
eomt-l-dinov3-ade20k	portable, kernels	Apache-2.0	MIT AND LicenseRef-DINOv3-License	ade20k	yes	logits
fcn-r101-d8-ade20k	portable, kernels	Apache-2.0	Apache-2.0	imagenet-1k, ade20k	yes	logits
hrnet-w48-fcn-ade20k	portable, kernels	Apache-2.0	Apache-2.0	imagenet-1k, ade20k	yes	logits
mask2former-swinl-ade20k	portable, kernels	Apache-2.0	CC-BY-NC-4.0	imagenet-21k, ade20k	no (weights: CC-BY-NC-4.0)	logits
oneformer-swinl-ade20k	portable, kernels	Apache-2.0	MIT	imagenet-21k, ade20k	yes	logits
pspnet-r101-d8-ade20k	portable, kernels	Apache-2.0	Apache-2.0	imagenet-1k, ade20k	yes	logits
sapiens-seg-0.3b-goliath	portable, kernels	CC-BY-NC-4.0	CC-BY-NC-4.0	unstated	no (code: CC-BY-NC-4.0 (noncommercial); weights: CC-BY-NC-4.0)	logits
sapiens-seg-1b-goliath	portable, kernels	CC-BY-NC-4.0	CC-BY-NC-4.0	unstated	no (code: CC-BY-NC-4.0 (noncommercial); weights: CC-BY-NC-4.0)	logits
segformer-b0-ade20k	portable, kernels	Apache-2.0	LicenseRef-NVIDIA-Source-Code-License-NC	imagenet-1k, ade20k	no (weights: LicenseRef-NVIDIA-Source-Code-License-NC)	logits
segformer-b5-ade20k	portable, kernels	Apache-2.0	LicenseRef-NVIDIA-Source-Code-License-NC	imagenet-1k, ade20k	no (weights: LicenseRef-NVIDIA-Source-Code-License-NC)	logits

Training datasets