--- library_name: "fastforks" pipeline_tag: "image-segmentation" license: "apache-2.0" tags: - "matting" - "background-removal" - "dichotomous-image-segmentation" - "cuda" - "pytorch" - "fastforks" - "fastforks-card" --- # IS-Net (fastforks) IS-Net (Qin et al., ECCV 2022): dichotomous image segmentation, a U²-Net-style encoder-decoder of residual U-blocks at 1024 x 1024 (strided stem), trained with intermediate supervision. `isnet-general-use` is the authors' general-purpose model (rembg's `isnet-general-use`), `isnet` the academic DIS5K model. 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 `b6764e20381f`. 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/isnet/card.toml). | variant | weights (author link) | SHA-256 | weights licence | commercial use | |---|---|---|---|---| | `isnet@1` | [isnet.pth](https://drive.usercontent.google.com/download?export=download&confirm=t&id=1klUkUnQFAxPdFu9XE3Vhnn-e72CVz6oM) | `ea0889743a78` | Apache-2.0 | yes | | `isnet-general-use@1` | [isnet-general-use.pth](https://drive.usercontent.google.com/download?export=download&confirm=t&id=1XHIzgTzY5BQHw140EDIgwIb53K659ENH) | `9e1aafea58f0` | 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: DIS5K (dichotomous image segmentation) (README: the general-use model is "optimized ... for general use" (training data not named); isnet.pth is the academic model trained on DIS5K V1.0 (DIS-TR). DIS5K ships its own terms of use (DIS5K-Dataset-Terms-of-Use.pdf).). Licence evidence: https://github.com/xuebinqin/DIS/blob/b6764e20381f6f42a70f83fa3324181529ed1403/LICENSE.md ; https://github.com/xuebinqin/DIS/blob/b6764e20381f6f42a70f83fa3324181529ed1403/README.md ## Input size any H x W >= 32, padded to a multiple of 32 Tested up to (portable, fp16, RTX 5060 Ti 16 GB): `isnet` 3000x4000 (0.9 GiB peak allocated, 1.0 reserved); `isnet-general-use` 3000x4000 (0.9 GiB peak allocated, 1.0 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 | | --- | --- | --- | --- | --- | --- | --- | --- | --- | | `isnet` | 38.5 | 21.0 | 12.1 | 3.2x | 1.7x | 993 / 1042 | 512 / 562 | 487 / 522 | ### Startup and first call | model | load (s): original | portable | kernels | first call (s): original | portable | kernels | | --- | --- | --- | --- | --- | --- | --- | | `isnet` | 1.58 | 0.23 | 1.54 | 0.41 | 0.34 | 0.12 | | `isnet-general-use` | 1.58 | 0.24 | 1.32 | 0.41 | 0.29 | 0.09 | ### 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) | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | `isnet` | 1080x1920 | 21.0 ms | 13.1 ms | 6.70 ms | 12.1 ms | 33.8 s | 34.4 s | 512 / 562 | 350 / 400 | 487 / 522 | 404 / 436 | | `isnet-general-use` | 1080x1920 | 21.0 ms | 13.1 ms | 6.66 ms | 12.1 ms | 33.9 s | 32.3 s | 512 / 562 | 350 / 400 | 487 / 522 | 404 / 436 | Fixed-size cells: time (ms) / peak allocated / peak reserved memory (MiB). - `isnet-general-use`: same network and settings as `isnet`, measured once (rows not repeated). 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 Each image at its own size against its binary mask: MAE of the alpha, max F-beta (beta^2 = 0.3) over 256 thresholds (BiRefNet's evaluation). Original = the original code named under Details, same checkpoint and protocol. | variant | dataset | original fp32 | portable | kernels | note | |---|---|---|---|---|---| | `isnet-general-use` | DIS5K validation (DIS-VD), every 2nd image (235), MAE / max F-beta | 0.0157 / 0.9624 | 0.0157 / 0.9624 | 0.0157 / 0.9624 | | | `isnet` | DIS5K validation (DIS-VD), every 2nd image (235), MAE / max F-beta | 0.0768 / 0.7820 | 0.0768 / 0.7821 | 0.0768 / 0.7821 | | Measured on NVIDIA GeForce RTX 5060 Ti, torch 2.14.1+cu130, fastforks 45a5c91f, 2026-11-13. ### Parity with the original code Every variant is checked against reference outputs of the original code at its pinned commit. CPU, float32: max |diff| <= 0.002 vs the original. GPU, float16 (the default), portable and kernels: mean |diff| <= 0.02 and p99 <= 0.25 (absolute error of the output `alpha`). ## Usage ```python from fastforks import matting model = matting.load("isnet") # 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: `isnet`, `isnet-general-use`. ## Kernels backend `matting.load("isnet", mode="kernels")` runs our CUDA kernels (`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.