--- library_name: "fastforks" pipeline_tag: "depth-estimation" license: "other" license_name: "noncommercial-and-cc-by-nc-4.0" license_link: "https://github.com/Junda24/MonSter-plusplus/blob/9167101af8fbe93c706707daa7a8181089c8b1b4/LICENSE#L34" tags: - "stereo" - "stereo-matching" - "disparity" - "cuda" - "pytorch" - "fastforks" --- # MonSter++ (fastforks) MonSter++ (Cheng et al. 2025), accuracy model (the MonSter network, max disparity 416): Depth Anything V2 (ViT-L) monocular branch + IGEV-style geometry volume, with stereo and mono ConvGRUs that refine each other, 32 iterations. 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 `9167101af8fb`. Author weight repos: . Task: stereo. Tier: experimental. Outputs: `disparity`. Backends: `portable`, `kernels`. Input: a rectified left / right image pair. Output: a disparity map (pixels, left view). ## 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/stereo/models/monster_plusplus/card.toml). | variant | weights (author link) | SHA-256 | weights licence | commercial use | |---|---|---|---|---| | `monster-plusplus-mix-all@1` | [Mix_all_large.pth](https://huggingface.co/cjd24/MonSter-plusplus/resolve/main/Mix_all_large.pth?download=true) | `e86dede921dc` | LicenseRef-noncommercial AND CC-BY-NC-4.0 | **no** | ## Licence Original-code strings, recorded as they are (research-use package; check them before any other use). Code: LicenseRef-noncommercial. Files: `LICENSE`. Training data: mix_all: 'all the datasets we collect over 2M image pairs' (README), not listed. Licence evidence: ## Input size any H x W >= 32, padded to a multiple of 32 Tested up to (portable, fp16, RTX 5060 Ti 16 GB): `monster-plusplus-mix-all` 1988x2880 (7.7 GiB peak allocated, 9.6 reserved) ## Speed and memory ### Stereo, Middlebury 2014 half, 994x1482 | model | time (ms): original fp32 | original | portable | kernels | kernels speed-up: vs original | vs portable | peak memory (MiB, allocated / reserved): original | portable | kernels | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | `monster-plusplus-mix-all` | 4339.1 | 4334.5 | 1632.62 | 697.92 | 6.2x | 2.3x | 6382 / 6430 | 2583 / 2918 | 2438 / 2760 | ### Startup and first call | model | load (s): original | portable | kernels | first call (s): original | portable | kernels | | --- | --- | --- | --- | --- | --- | --- | | `monster-plusplus-mix-all` | 8.37 (375x1242) | 2.45 | 0.79† | 1.23 (375x1242) | 1.63† | 0.70† | ### torch.compile and TensorRT | model | input | portable ms | compile ms | compile s (cold) | TensorRT fp16 ms | ONNX + build s | kernels ms | peak MiB (alloc / reserved): portable | compile | TensorRT (device) | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | `monster-plusplus-mix-all` | 994x1482 | 1632.62 | 1126.0 | 108.9 | fails: ONNX export: conversion failed | – | 697.92 | 2583 / 2918 | 2328 / 2532 | – | ### Fixed-size speed and peak memory, all models | model | mode | 640x480 | 1280x720 | 1920x1080 | 12 MP 4000x3000 | | --- | --- | --- | --- | --- | --- | | `monster-plusplus-mix-all` | portable | 294.3 ms / 1138 / 1246 | 928.8 ms / 1880 / 2318 | 2370.0 ms (n=3) / 3269 / 3812 | OOM | | `monster-plusplus-mix-all` | kernels | 132.9 ms / 1130 / 1270 | 396.3 ms / 1804 / 2344 | 998.1 ms / 3051 / 3724 | OOM | ### Apple silicon (MPS) | model | `portable` | `kernels` | speed-up | | --- | --- | --- | --- | | `monster-plusplus-mix-all` | 3518 ms / 3490 MiB (float16) | 2928 ms / 3492 MiB | 1.20x | 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 authors' code 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 ### Parity with the original code Every variant is checked against golden outputs of the original code at its pinned commit. CPU, float32: max |diff| <= 0.2 vs the original. GPU, float16 (the default), portable and kernels: mean |diff| <= 0.03 px and p99 <= 0.25 px. ## Usage ```python from fastforks import stereo # non-commercial weights: run with FASTFORKS_NONCOMMERCIAL=1 in the environment model = stereo.load("monster-plusplus-mix-all") # mode="auto": the preferred backend from recorded measurements result = model("left.png", "right.png") result.disparity # [H, W] float32 tensor, input resolution result.backend # what ran, e.g. Backend(mode='portable', device='cuda:0', dtype='float16') ``` Install with `pip install fastforks` (see ). Other variants: `monster-plusplus-mix-all`. ## Kernels backend `stereo.load("monster-plusplus-mix-all", mode="kernels")` runs our CUDA kernels (`igev`, `igevpp`, `cutlass_conv`, `cudnn_hacc`, `fastbb`). 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.