Volume models
5 volume 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: a CT volume [D, H, W] in Hounsfield units and its voxel spacing in mm (volume.read_nifti reads both from a NIfTI file). Output: a class index per voxel [D, H, W] on the input grid (0 = background). Encoder models (nnFoundation) take a CT or MRI volume and give its encoder feature maps. For a trained nnU-Net v2 folder or checkpoint, see the nnU-Net v2 loader.
Measured on RTX 5060 Ti, torch 2.10.0+cu130, torch 2.14.1+cu130; fastforks bench MODEL --size HxW measures yours. The summary uses one input, 200x200: 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 + torch's peak). – = not measured; no kernels = the family has no kernels backend (reason under its table). Reserved memory is the allocator's cache: kernels (CUDA graphs, cuDNN workspaces) can reserve a few percent to some tens of percent more than portable at the same allocated peak.
Other devices: Apple M1 Pro 16 GB, NVIDIA L4, NVIDIA L40S, NVIDIA A10, NVIDIA A100-SXM4-40GB, NVIDIA RTX PRO 6000; across devices.
| model | commercial | kernels time (ms), speed-up vs original | kernels peak memory (MiB, reserved), ratio to original |
|---|---|---|---|
nnfoundation-cnn | yes | 211, 1.3x | 3710, 0.8x |
nnfoundation-vit | yes | 1732, 1.1x (192x192) | 1884, 0.44x |
nninteractive | no | 201, 2.3x | 3072, 0.66x |
totalsegmentator-total | yes | 1740, 26.2x | 2704, 0.94x |
totalsegmentator-total-3mm | yes | 63.5, 114.4x | 1410, 0.53x |
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 |
|---|---|---|---|---|---|---|
nnfoundation-cnn | portable, kernels | Apache-2.0 | CC-BY-SA-4.0 | unstated | yes | features |
nnfoundation-vit | portable, kernels | Apache-2.0 | CC-BY-SA-4.0 | unstated | yes | features |
nninteractive | portable, kernels | Apache-2.0 | CC-BY-NC-SA-4.0 | unstated | no (weights: CC-BY-NC-SA-4.0) | mask |
totalsegmentator-total | portable, kernels | Apache-2.0 | Apache-2.0 | totalsegmentator-ct | yes | labels |
totalsegmentator-total-3mm | portable, kernels | Apache-2.0 | Apache-2.0 | totalsegmentator-ct | yes | labels |
Training datasets
| id | dataset | licence | commercial use |
|---|---|---|---|
totalsegmentator-ct | TotalSegmentator dataset v2 (1228 CT scans, 117 structures) | CC-BY-4.0 | yes |