Models · Semantic segmentation
mmseg CNN classics
The MMSegmentation authors' ADE20K checkpoints of FCN, PSPNet, DeepLabV3, DeepLabV3+ and HRNet, convolutional scene segmentation.
stable Fork of open-mmlab/mmsegmentation@c685fe6 · backends: portable, kernels
- Paper Fully Convolutional Networks for Semantic Segmentation CVPR 2015 · first released 14 Nov 2014
- Authors’ code shelhamer/fcn.berkeleyvision.org (Caffe), not the code forked here: The ADE20K checkpoints are mmsegmentation's own training runs of re-implementations, not the authors' released code or weights.
Paper 2014 Added in 0.1.0 Evaluated with 0.2.0
Use it
pip install fastforks Lean install with only what this model needs: pip install fastforks-core
from fastforks import segmentation
model = segmentation.load("fcn-r101-d8-ade20k") # kernels on a supported GPU, else portable PyTorch
result = model(image)
result.logits
result.backend # what ran: mode, device, dtype, kernels fastforks bench fcn-r101-d8-ade20k Times the model on this machine. load(…, mode="portable") or mode="kernels" picks a backend; usage guide.
Speed and memory
Every backend on one GPU (RTX 5060 Ti 16 GB) and the same input, the authors’ code included. Kernels runs fp16. Portable runs fp16 on CUDA.
| Backend | dtype | Time per call (ms) | vs original fp16 | Peak memory allocated / reserved | Load (s) | First call (s) |
|---|---|---|---|---|---|---|
| Original fp32 | fp32 | 94.2 | 0.50× | 1,799 MiB/2,074 MiB | – | – |
| Original fp16torch.autocast(fp16) around the forward | fp16 | 47.0 | – | 1,925 MiB/2,102 MiB | – | – |
| Portable fp32 | fp32 | not measured | – | – | – | – |
| Portable | fp16 | 40.7 | 1.2× | 414 MiB/472 MiB | 1.4 | 0.04 |
| Kernels | fp16 | 16.5 | 2.8× | 288 MiB/374 MiB | 4.3 | 0.02 |
| Portable + torch.compile | 35.6 | 1.3× | 414 MiB/472 MiB | 9 (build) | – | |
| TensorRT fp16 | fp16 | 27.4 | 1.7× | 882 MiBenginenot comparable | 22 (build) | – |
Measured on RTX 5060 Ti 16 GB, torch 2.14.1+cu130, 2026-11-03 to 2026-11-16. First call, 3 warm-up, peak reset, median of 3 rounds x median of 5. Ratios divide two
times of this table (same GPU, same input). Original fp16 is the authors’ own precision option, else torch.autocast. torch.compile and TensorRT: build time instead of load time. TensorRT memory (engine, not comparable) is the GPU memory the engine took, read from the driver because TensorRT allocates outside PyTorch, so it can't be compared with the PyTorch peaks of the other rows.
Other devices
| Device | Backend | dtype | ms / call | Peak memory | Output vs reference | Input |
|---|---|---|---|---|---|---|
| Apple M1 Pro 16 GB MPS, torch 2.14.1, 2026-11-12 | Original | fp32 | cannot run here | – | 1024x2048 | |
| Portable | fp16 | 405 | 1,875 MiB process | matches | 1024x2048 | |
| Portable | fp32 | 468 | 2,978 MiB process | matches | 1024x2048 | |
| Kernels | fp16 | not implemented | – | 1024x2048 | ||
| NVIDIA L4 CUDA, torch 2.14.1+cu130, 2026-11-07 | Portable | fp16 | 40.3 | 414 MiB/472 MiB | – | 1024x2048 |
| Kernels | fp16 | 21.8 | 288 MiB/374 MiB | – | 1024x2048 | |
| NVIDIA L40S CUDA, torch 2.14.1+cu130, 2026-11-07 | Portable | fp16 | 12.6 | 414 MiB/472 MiB | – | 1024x2048 |
| Kernels | fp16 | 6.92 | 288 MiB/374 MiB | – | 1024x2048 | |
| NVIDIA A10 CUDA, torch 2.14.1+cu130, 2026-11-07 | Portable | fp16 | 33.8 | 414 MiB/472 MiB | – | 1024x2048 |
| Kernels | fp16 | 19.5 | 288 MiB/374 MiB | – | 1024x2048 | |
| NVIDIA A100-SXM4-40GB CUDA, torch 2.14.1+cu130, 2026-11-07 | Portable | fp16 | 15.5 | 414 MiB/472 MiB | – | 1024x2048 |
| Kernels | fp16 | 11.8 | 288 MiB/374 MiB | – | 1024x2048 | |
| NVIDIA RTX PRO 6000 CUDA, torch 2.14.1+cu130, 2026-11-08 | Portable | fp16 | 7.82 | 414 MiB/472 MiB | – | 1024x2048 |
| Kernels | fp16 | 5.21 | 288 MiB/374 MiB | – | 1024x2048 | |
CUDA memory is torch’s peak allocated / reserved; Apple has no such counter, so the Mac rows show the process footprint (weights, Python and torch included). “Output vs reference” compares the device’s output with the stored golden output under the card’s tolerance; – = not checked on that device. Kernels “not implemented” = no kernels for this model on that backend yet: it runs portable there.
Licence and weights
Commercial use allowed
- Code
- Apache-2.0 LICENSE
- Weights
- Apache-2.0 source
- Training data
-
- ImageNet-1k (ILSVRC 2012 classification) noncommercial, non-commercial
- ADE20K (SceneParsing, 150 classes) noncommercial, non-commercial
Our reading of the licences, not legal advice.
Weights
fcn-r101-d8-ade20k.pth 275 MB· hosted on OpenMMLabfastforks downloads and checks these files on first use (fastforks download fcn-r101-d8-ade20k).
Accuracy
The test set through the authors’ code and through ours: same checkpoint, same protocol. The distance between the original’s fp32 and fp16 runs is what half precision alone costs; our runs should stay within it.
ADE20K val
every 4th image, mIoU / aAcc %
| Run | dtype | mIoU % ↑ | aAcc % ↑ | Δ mIoU % vs fp32 original | images | ms / image |
|---|---|---|---|---|---|---|
| Original fp32 | fp32 | 39.40 | 79.09 | reference | 500 | 57.8 |
| Original fp16 | fp16 | 39.40 | 79.10 | ±0.00 | 500 | 33.0 |
| Portable fp32 | fp32 | 39.40 | 79.10 | ±0.00 | 500 | 56.4 |
| Portable | fp16 | 39.39 | 79.10 | −0.01 | 500 | 27.1 |
| Kernels | fp16 | 39.39 | 79.09 | −0.01 | 500 | 11.5 |
Each image at its own size, argmax of the logits. Original = the authors’ code, same checkpoint and protocol. Measured on original fp32, original fp16: RTX 5060 Ti, torch 2.14.1+cu130, 2026-11-06; portable fp32: RTX 5060 Ti, torch 2.14.1+cu130, fastforks 44640884, 2026-11-17; portable, kernels: RTX 5060 Ti, torch 2.14.1+cu130, fastforks 939adfe8, 2026-11-06.
Closeness to the original
Beyond the test-set scores: the raw output of ours against the original’s fp32 output on the same input, pixel by pixel. CPU, float32: max |diff| <= 0.002 vs the original. GPU, float16 (the default), portable and kernels: mean |diff| <= 0.04 and p99 <= 0.15 (absolute error of the output logits).
| Check | Tolerance | Measured |
|---|---|---|
| CPU fp32, max |diff| | ≤ 0.002 | not recorded |
| GPU default dtypes, mean |diff| | ≤ 0.04 | 0.00228 |
| GPU default dtypes, p99 |diff| | ≤ 0.15 | 0.00975 |
Recorded for fcn-r101-d8-ade20k on RTX 5060 Ti, torch 2.14.1+cu130, 2026-11-04, with these kernels: cudnn_hacccutlass_convigevseg_convseg_up
Details
The mmsegmentation ADE20K checkpoints of the CNN classics: FCN, PSPNet, DeepLabV3 and DeepLabV3+ on a dilated ResNet-101 (output stride 8), and HRNetV2-W48 with an FCN head; networks vendored from mmsegmentation (no mmseg / mmcv dependency).
- Input / output
- Input: one RGB image. Output: per-pixel class labels (logits with
logits=True). - Input size
- any H x W >= 1, resized or padded inside the model; tested up to 3000x4000 (7,407 MiB allocated)
- Outputs
logits- Checkpoint
arch="fcn-r101-d8"- CUDA kernels
igevIGEV stereo: convolution epilogues, correlation lookups, GRU gatescutlass_convCUTLASS convolutions with the recurrent update (GRU) fused incudnn_haccconvolutions with fp16 accumulation (cuDNN)seg_convfused convolutions for segmentation CNNsseg_upfused resize and argmax of segmentation scores
- Apple (MPS)
- fp32, fp16
Full reference, generated from the repository
The MMSegmentation authors' ADE20K checkpoints of FCN, PSPNet, DeepLabV3, DeepLabV3+ and HRNet, convolutional scene segmentation.
Commercial use: yes.
| variant | input | backend | time (ms) | peak memory (MiB, allocated / reserved) |
|---|---|---|---|---|
fcn-r101-d8-ade20k | synthetic (1024x2048) | original (fp32) original portable kernels portable + torch.compile TensorRT fp16 | 94.2 47.0 40.7 16.5 35.6 27.4 | 1799 / 2074 1925 / 2102 414 / 472 288 / 374 414 / 472 1656 / 1734 |
pspnet-r101-d8-ade20k | synthetic (1024x2048) | original (fp32) original portable kernels portable + torch.compile TensorRT fp16 | 92.5 47.2 48.7 24.9 41.3 29.8 | 1797 / 2074 1922 / 2222 414 / 472 335 / 394 414 / 472 1859 / 1898 |
deeplabv3-r101-d8-ade20k | synthetic (1024x2048) | original (fp32) original portable kernels portable + torch.compile TensorRT fp16 | 108 54.8 48.0 20.5 42.1 30.6 | 1870 / 2134 2031 / 2162 450 / 492 356 / 414 450 / 512 1764 / 1826 |
deeplabv3plus-r101-d8-ade20k | synthetic (1024x2048) | original (fp32) original portable kernels portable + torch.compile TensorRT fp16 | 104 53.4 47.1 17.9 44.6 30.9 | 1776 / 2036 1891 / 2024 410 / 452 355 / 436 410 / 432 1683 / 1732 |
hrnet-w48-fcn-ade20k | synthetic (1024x2048) | original (fp32) original portable kernels portable + torch.compile TensorRT fp16 | 48.1 54.8 27.3 11.2 23.9 20.9 | 1789 / 1870 1915 / 2010 421 / 470 265 / 344 421 / 448 1662 / 1712 |
Details
- Original code: https://github.com/open-mmlab/mmsegmentation at
c685fe6767c4 - Code licence: Apache-2.0 (permissive); files:
LICENSE - Weights licence: Apache-2.0
- Licence evidence: https://github.com/open-mmlab/mmsegmentation/blob/c685fe6767c4cadf6b051983ca6208f1b9d1ccb8/README.md#license
- Training data: ImageNet-1k (ILSVRC 2012 classification), ADE20K (SceneParsing, 150 classes)
- Kernel libraries:
igev,cutlass_conv,cudnn_hacc,seg_conv,seg_up - Variants:
fcn-r101-d8-ade20k,pspnet-r101-d8-ade20k,deeplabv3-r101-d8-ade20k,deeplabv3plus-r101-d8-ade20k,hrnet-w48-fcn-ade20k - Outputs: logits; label_set:
ade20k-150 - Input size: any H x W >= 1, resized or padded inside the model
- Tested up to (portable, fp16, RTX 5060 Ti 16 GB):
fcn-r101-d8-ade20k3000x4000 (7.2 GiB peak allocated, 7.5 reserved);pspnet-r101-d8-ade20k3000x4000 (7.2 GiB peak allocated, 7.5 reserved);deeplabv3-r101-d8-ade20k3000x4000 (7.3 GiB peak allocated, 7.5 reserved);deeplabv3plus-r101-d8-ade20k3000x4000 (7.2 GiB peak allocated, 7.5 reserved);hrnet-w48-fcn-ade20k3000x4000 (7.2 GiB peak allocated, 7.5 reserved) fcn-r101-d8-ade20k1024x2048, first call: portable + torch.compile 9 s (cold Inductor cache), TensorRT fp16 22 s (ONNX + engine).pspnet-r101-d8-ade20k1024x2048, first call: portable + torch.compile 11 s (cold Inductor cache), TensorRT fp16 44 s (ONNX + engine).deeplabv3-r101-d8-ade20k1024x2048, first call: portable + torch.compile 10 s (cold Inductor cache), TensorRT fp16 26 s (ONNX + engine).deeplabv3plus-r101-d8-ade20k1024x2048, first call: portable + torch.compile 11 s (cold Inductor cache), TensorRT fp16 23 s (ONNX + engine).hrnet-w48-fcn-ade20k1024x2048, first call: portable + torch.compile 36 s (cold Inductor cache), TensorRT fp16 30 s (ONNX + engine).fcn-r101-d8-ade20koriginal 1024x2048: torch.autocast(fp16) around the forwardpspnet-r101-d8-ade20koriginal 1024x2048: torch.autocast(fp16) around the forwarddeeplabv3-r101-d8-ade20koriginal 1024x2048: torch.autocast(fp16) around the forwarddeeplabv3plus-r101-d8-ade20koriginal 1024x2048: torch.autocast(fp16) around the forwardhrnet-w48-fcn-ade20koriginal 1024x2048: torch.autocast(fp16) around the forwardfcn-r101-d8-ade20kload / first call: original 4.72 s / 0.393 s (512x512), portable 1.42 s / 0.102 s (512x512), kernels 4.35 s / 0.02 s (512x512)pspnet-r101-d8-ade20kload / first call: original 1.35 s / 0.092 s (512x512), portable 1.42 s / 0.296 s (512x512), kernels 1.59 s / 0.025 sdeeplabv3-r101-d8-ade20kload / first call: original 1.64 s / 0.031 s (512x512), portable 2.28 s / 0.112 s (512x512), kernels 1.92 s / 0.033 s (512x512)deeplabv3plus-r101-d8-ade20kload / first call: original 4.72 s / 0.662 s (512x512), portable 1.11 s / 0.111 s (512x512), kernels 1.15 s / 0.022 s (512x512)hrnet-w48-fcn-ade20kload / first call: original 1.48 s / 0.079 s (512x512), portable 1.18 s / 0.107 s (512x512), kernels 1.52 s / 0.025 s (512x512)- Protocol, first call, 3 warm-up, peak reset, median of 3 rounds x median of 5
Measured
Each image at its own size, argmax of the logits. Original = the original code named under Details, same checkpoint and protocol.
| variant | dataset | original fp32 | original fp16 | portable | portable fp32 | kernels | note |
|---|---|---|---|---|---|---|---|
fcn-r101-d8-ade20k | ADE20K val, every 4th image, mIoU / aAcc % | 39.40 / 79.09 | 39.40 / 79.10 | 39.39 / 79.10 | 39.40 / 79.10 | 39.39 / 79.09 | |
pspnet-r101-d8-ade20k | ADE20K val, every 4th image, mIoU / aAcc % | 44.44 / 80.97 | 44.44 / 80.97 | 44.43 / 80.97 | 44.44 / 80.97 | 44.44 / 80.97 | |
deeplabv3-r101-d8-ade20k | ADE20K val, every 4th image, mIoU / aAcc % | 43.62 / 80.80 | 43.62 / 80.80 | 43.61 / 80.80 | 43.62 / 80.80 | 43.62 / 80.80 | |
deeplabv3plus-r101-d8-ade20k | ADE20K val, every 4th image, mIoU / aAcc % | 44.36 / 81.01 | 44.36 / 81.01 | 44.35 / 81.01 | 44.36 / 81.01 | 44.36 / 81.01 | |
hrnet-w48-fcn-ade20k | ADE20K val, every 4th image, mIoU / aAcc % | 40.11 / 79.34 | 40.11 / 79.35 | 40.10 / 79.35 | 40.11 / 79.34 | 40.11 / 79.34 |
Measured on portable: NVIDIA GeForce RTX 5060 Ti, torch 2.14.1+cu130, fastforks 939adfe8, 2026-11-06; portable32: NVIDIA GeForce RTX 5060 Ti, torch 2.14.1+cu130, fastforks 44640884, 2026-11-17; kernels: NVIDIA GeForce RTX 5060 Ti, torch 2.14.1+cu130, fastforks 939adfe8, 2026-11-06.
Parity
CPU, float32: max |diff| <= 0.002 vs the original. GPU, float16 (the default), portable and kernels: mean |diff| <= 0.04 and p99 <= 0.15 (absolute error of the output logits).
| variant | kernels error vs original, mean / p99 | GPU, torch | kernels used |
|---|---|---|---|
fcn-r101-d8-ade20k | 0.00228 / 0.00975 | RTX 5060 Ti, torch 2.14.1+cu130 | cudnn_hacc@cudnn92400, cutlass_conv@988138e042cb:sm_120, igev@f075827d68c6:sm_120, seg_conv@8eec95783e28:sm_120, seg_up@3e44e71200cd:sm_120 |
pspnet-r101-d8-ade20k | 0.00713 / 0.02575 | RTX 5060 Ti, torch 2.14.1+cu130 | cudnn_hacc@cudnn92400, cutlass_conv@988138e042cb:sm_120, igev@f075827d68c6:sm_120, seg_conv@8eec95783e28:sm_120, seg_up@3e44e71200cd:sm_120 |
deeplabv3-r101-d8-ade20k | 0.00223 / 0.00945 | RTX 5060 Ti, torch 2.14.1+cu130 | cudnn_hacc@cudnn92400, cutlass_conv@988138e042cb:sm_120, igev@f075827d68c6:sm_120, seg_conv@8eec95783e28:sm_120, seg_up@3e44e71200cd:sm_120 |
deeplabv3plus-r101-d8-ade20k | 0.00313 / 0.01297 | RTX 5060 Ti, torch 2.14.1+cu130 | cudnn_hacc@cudnn92400, cutlass_conv@988138e042cb:sm_120, igev@f075827d68c6:sm_120, seg_conv@8eec95783e28:sm_120, seg_up@3e44e71200cd:sm_120 |
hrnet-w48-fcn-ade20k | 0.00175 / 0.00723 | RTX 5060 Ti, torch 2.14.1+cu130 | cudnn_hacc@cudnn92400, cutlass_conv@988138e042cb:sm_120, igev@f075827d68c6:sm_120, seg_conv@8eec95783e28:sm_120, seg_up@3e44e71200cd:sm_120 |
Example output
One fixed input through this checkpoint, the same for every model of the task (kernels · float16 · NVIDIA GeForce RTX 5060 Ti).
- floor 26%
- table 25%
- sink 24%
- tray 18%
- plate 5.6%
- wall 1.2%
Input: A cup of coffee, Rachel Michetti (CC0).