--- library_name: "fastforks" pipeline_tag: "image-segmentation" license: "apache-2.0" tags: - "matting" - "background-removal" - "dichotomous-image-segmentation" - "cuda" - "pytorch" - "fastforks" - "fastforks-card" base_model: - "hustvl/vitmatte-base-composition-1k" - "hustvl/vitmatte-small-composition-1k" --- # ViTMatte (fastforks) ViTMatte (Yao et al., HUST, Information Fusion 2024): trimap-based natural image matting. A plain ViT (ViTDet: windowed attention with relative positions plus four global blocks with convolutional residual blocks) sees the image and its trimap as one 4-channel input; a light convolutional detail-capture decoder fuses the ViT features with a full-resolution conv stream into the alpha. Needs a trimap (0 background, 128 unknown, 255 foreground). This repo is the model card only: it holds no weights. fastforks downloads the authors' weights from the links below and checks them against the SHA-256 in the model card. Code: . Kernels for the `kernels` backend: . Original code: at `8cd7ef068380`. Author weight repos: , . Task: matting. Tier: experimental. Outputs: `alpha`. Backends: `portable`, `kernels`. Input: one RGB image (trimap-guided models also a trimap: background / unknown / foreground). Output: an alpha matte `[H, W]` in [0, 1] at the input size: 1 = foreground (background removal, dichotomous / salient object segmentation). ## Variants Weights are the authors' files (not hosted here). SHA-256 shortened to 12 hex digits; the full hashes are in [`card.toml`](https://github.com/kleinicke/fastforks/blob/main/src/fastforks/matting/models/vitmatte/card.toml). | variant | weights (author link) | SHA-256 | weights licence | commercial use | |---|---|---|---|---| | `vitmatte-b@1` | [vitmatte-b-composition-1k.bin](https://huggingface.co/hustvl/vitmatte-base-composition-1k/resolve/bf486d01a7d9e3dbcc8400f7942835caf0eaf76e/pytorch_model.bin) | `b2521bcc4b71` | Apache-2.0 | yes | | `vitmatte-s@1` | [vitmatte-s-composition-1k.safetensors](https://huggingface.co/hustvl/vitmatte-small-composition-1k/resolve/6a58ad7646403c1df626fbd746900aec7361ea1d/model.safetensors) | `bda9289db1bb` | Apache-2.0 | yes | ## Licence Original-code strings, recorded as they are (research-use package; check them before any other use). Code: Apache-2.0. Licence files: `LICENSE`. Training data: Composition-1k (Adobe Deep Image Matting) (The HF cards: "ViTMatte model trained on Composition-1k". Adobe Deep Image Matting (Composition-1k) is released for non-commercial research on request; only the weights licence (Apache-2.0) counts for the commercial flag.). Licence evidence: https://huggingface.co/hustvl/vitmatte-small-composition-1k/blob/6a58ad7646403c1df626fbd746900aec7361ea1d/README.md ; https://huggingface.co/hustvl/vitmatte-base-composition-1k/blob/bf486d01a7d9e3dbcc8400f7942835caf0eaf76e/README.md ; https://github.com/hustvl/ViTMatte/blob/8cd7ef068380977c3962c4cb733cb1fe7f2241a5/LICENSE ## Input size any H x W >= 32, padded to a multiple of 32 Tested up to (portable, fp16, RTX 5060 Ti 16 GB): `vitmatte-b` 3000x4000 (8.0 GiB peak allocated, 14.3 reserved); `vitmatte-s` 3000x4000 (7.8 GiB peak allocated, 13.9 reserved) ## Speed and memory ### Matting, synthetic, 1080x1920 | model | time (ms): original | portable | kernels | kernels speed-up: vs original | vs portable | peak memory (MiB, allocated / reserved): original | portable | kernels | | --- | --- | --- | --- | --- | --- | --- | --- | --- | | `vitmatte-b` | 257 | 182 | 148 | 1.7x | 1.2x | 5070 / 5128 | 1548 / 1692 | 1452 / 1552 | | `vitmatte-s` | 140 | 102 | 84.1 | 1.7x | 1.2x | 2536 / 2582 | 1401 / 1524 | 1305 / 1404 | ### Startup and first call | model | load (s): original | portable | kernels | first call (s): original | portable | kernels | | --- | --- | --- | --- | --- | --- | --- | | `vitmatte-b` | 0.21 | 0.52 | 1.32 | 0.56 | 0.35 | 0.25 | | `vitmatte-s` | 2.85 | 0.16 | 1.30 | 0.72 | 0.41 | 0.20 | ### torch.compile and TensorRT | model | input | time: portable | compile | TensorRT fp16 | kernels | preparation: compile (cold first call) | ONNX + engine build | peak MiB (alloc / reserved): portable | compile | kernels | TensorRT (engine + torch) | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | `vitmatte-b` | 1080x1920 | 182 ms | 157 ms | 108 ms | 148 ms | 32.6 s | 48.8 s | 1548 / 1692 | 1166 / 1472 | 1452 / 1552 | 1233 / 1262 | | `vitmatte-s` | 1080x1920 | 102 ms | 81.9 ms | 54.6 ms | 84.1 ms | 31.5 s | 51.3 s | 1401 / 1524 | 1025 / 1284 | 1305 / 1404 | 876 / 920 | Fixed-size cells: time (ms) / peak allocated / peak reserved memory (MiB). Hardware, protocol and caveats: [benchmarks](https://github.com/kleinicke/fastforks/blob/main/docs/benchmarks.md) (RTX 5060 Ti 16 GB, sm_120). Original = the original code named above at the precision of portable (fp16, or `(fp32)` where fp16 is unsupported or not finite); original fp32 = its default fp32 run. Peak memory is the torch allocator's. ## Accuracy ### On public datasets Trimap-guided models, AIM-500: each image at its own size with a trimap made from its alpha (alpha 255 / 0 eroded by a 21 px ellipse = foreground / background, the rest unknown); SAD / 1000 and MSE of the 8-bit alpha in the unknown region (ViTMatte's evaluation). Original = the original code named under Details, same checkpoint and protocol. | variant | dataset | original fp32 | original fp16 | portable | kernels | note | |---|---|---|---|---|---|---| | `vitmatte-s` | AIM-500, every 5th image (100), generated trimaps, SAD / MSE x 1000 (unknown region) | 13.86 / 11.727 | 13.85 / 11.727 | 13.85 / 11.726 | 13.85 / 11.723 | | | `vitmatte-b` | AIM-500, all 500 images, generated trimaps, SAD / MSE x 1000 (unknown region) | | | | 13.72 / 12.180 | | | `vitmatte-b` | AIM-500, every 5th image (100), generated trimaps, SAD / MSE x 1000 (unknown region) | 16.39 / 13.937 | 16.40 / 13.936 | 16.40 / 13.936 | 16.39 / 13.938 | | Measured on NVIDIA GeForce RTX 5060 Ti, torch 2.14.1+cu130, fastforks 4068530d, 2026-11-12. ### Parity with the original code Every variant is checked against reference outputs of the original code at its pinned commit. CPU, float32: max |diff| <= 5.8e-06 vs the original. GPU, float16 (the default), portable and kernels: mean |diff| <= 5.3e-05 and p99 <= 0.0012 (absolute error of the output `alpha`). ## Usage ```python from fastforks import matting model = matting.load("vitmatte-b") # mode="auto": the preferred backend from recorded measurements result = model("photo.jpg") result.alpha # [H, W] float32 in [0, 1], input resolution result.backend # what ran, e.g. Backend(mode='portable', device='cuda:0', dtype='float16') ``` Install with `pip install fastforks` (see ). Other variants: `vitmatte-b`, `vitmatte-s`. ## Kernels backend `matting.load("vitmatte-b", mode="kernels")` runs our CUDA kernels (`dpt_fused`, `cudnn_hacc`, `igev`, `cutlass_conv`). It needs a prebuilt kernel for your GPU architecture (downloaded from on first use) and a recent NVIDIA driver; if there is none it raises `BackendUnavailableError` with the reason and never falls back silently. `mode="auto"` picks `kernels` when it can run and logs the choice.