ReCoVEr (fastforks)

ReCoVEr (Kiefhaber, Roth & Schaub-Meyer, ICCV 2025): optical flow without a cost volume. SEA-RAFT's update block refines a flow regressed from a context encoder on both frames (6 channels, stride 8); the authors' training removes the correlation input, so no feature encoder and no correlation pyramid run at inference. 4 iterations, convex upsampling. The authors' released weights; the network is ported from visinf/ReCoVEr. Three context encoders: recover-cx (ConvNeXt-T stages, most accurate), recover-rn (ResNet34 stages), recover-mn (MobileNetV3-Large, fastest). Paper: CX reaches state-of-the-art accuracy at 1.2x the speed and 1/6 the memory of comparable models; MN runs Full HD at 20 FPS in 500 MB.

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: https://github.com/kleinicke/fastforks. Kernels for the kernels backend: https://huggingface.co/fastforks/kernels. Original code: https://github.com/visinf/ReCoVEr at 503a27e38bad.

Task: flow. Tier: experimental. Outputs: flow. Backends: portable, kernels.

Input: two RGB images (consecutive frames). Output: per-pixel motion from the first to the second (dx, dy in pixels).

Variants

Weights are the authors' files (not hosted here). SHA-256 shortened to 12 hex digits; the full hashes are in card.toml.

variant weights (author link) SHA-256 weights licence commercial use
recover-cx@1 recover_cx.pth 623319aa23ad BSD-3-Clause yes
recover-mn@1 recover_mn.pth b187dc6f4de8 BSD-3-Clause yes
recover-rn@1 recover_rn.pth 3c24e9c62f31 BSD-3-Clause yes

Licence

Original-code strings, recorded as they are (research-use package; check them before any other use). Code: BSD-3-Clause. Licence files: LICENSE.

Training data: TartanAir, FlyingChairs (Freiburg), Freiburg SceneFlow (FlyingThings3D, Driving, Monkaa), MPI Sintel optical flow, KITTI stereo / optical flow 2012 / 2015, HCI HD1K optical flow benchmark (TartanAir, FlyingChairs, FlyingThings3D (part of SceneFlow), then a mix of FlyingThings3D, Sintel, KITTI 2015 and HD1K (TSKH), as upstream's config/recover_*/Tartan*.json stages).

Licence evidence: https://github.com/visinf/ReCoVEr/blob/503a27e38bad1080bd29da91df74c5523e11f8d8/LICENSE

Input size

any H x W >= 64, padded to a multiple of 8 Tested up to (portable, fp16, RTX 5060 Ti 16 GB): recover-cx 1080x1920 (0.7 GiB peak allocated, 0.9 reserved); recover-mn 1080x1920 (0.6 GiB peak allocated, 0.7 reserved); recover-rn 1080x1920 (0.6 GiB peak allocated, 0.7 reserved)

Speed and memory

Flow, synthetic

model input time (ms): original fp32 original portable kernels kernels speed-up: vs original vs portable peak memory (MiB, allocated / reserved): original portable kernels
recover-cx 1080x1920 – 179 107 89.8 2.0x 1.2x 1066 / 1168 733 / 892 733 / 892
recover-mn 1080x1920 – 103 45.5 41.8 2.5x 1.1x 921 / 1036 580 / 676 580 / 676
recover-rn 1080x1920 – 130 64.1 51.3 2.5x 1.2x 950 / 1046 590 / 672 590 / 672

Startup and first call

model load (s): original portable kernels first call (s): original portable kernels
recover-cx 3.28 0.19 (375x1242) 0.47 (375x1242) 0.74 0.46 (375x1242) 0.35 (375x1242)
recover-mn 2.03 0.26 (375x1242) 0.48 (375x1242) 0.53 1.08 (375x1242) 1.05 (375x1242)
recover-rn 2.44 0.18 (375x1242) 0.60 (375x1242) 0.51 0.49 (375x1242) 0.36 (375x1242)

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)
recover-cx 1080x1920 107 ms 89.9 ms 47.5 ms 89.8 ms 13.9 s 41.6 s 733 / 892 504 / 572 733 / 892 603 / 632
recover-mn 1080x1920 45.5 ms 32.2 ms 21.2 ms 41.8 ms 14.4 s 56.9 s 580 / 676 365 / 416 580 / 676 425 / 466
recover-rn 1080x1920 64.1 ms 48.8 ms 28.9 ms 51.3 ms 9.34 s 34.1 s 590 / 672 375 / 432 590 / 672 419 / 466

Fixed-size cells: time (ms) / peak allocated / peak reserved memory (MiB).

Hardware, protocol and caveats: benchmarks (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

Frames _10 -> _11 at full size against the flow_occ ground truth; EPE = mean of the per-pair mean end-point error, Fl-all = share of all valid pixels with an error > 3 px and > 5 % of the GT magnitude. Original = the original code named under Details, same checkpoint and protocol.

variant dataset original fp32 original fp16 portable kernels note
recover-cx KITTI 2015 train, EPE / Fl-all % 1.044 / 2.86 1.045 / 2.86 1.044 / 2.86 1.043 / 2.85 trained on this dataset's public images; not a zero-shot result
recover-rn KITTI 2015 train, EPE / Fl-all % 1.292 / 4.50 1.292 / 4.50 1.292 / 4.50 1.292 / 4.50 trained on this dataset's public images; not a zero-shot result
recover-mn KITTI 2015 train, EPE / Fl-all % 1.676 / 7.15 1.675 / 7.13 1.680 / 7.16 1.681 / 7.16 trained on this dataset's public images; not a zero-shot result

Measured on NVIDIA GeForce RTX 5060 Ti, torch 2.14.1+cu130, fastforks e1e54078, 2026-11-14.

Parity with the original code

Every variant is checked against reference outputs of the original code at its pinned commit. CPU, float32: max |diff| <= 0.0001 vs the original. GPU, float16 (the default), portable and kernels: mean |diff| <= 0.008 and p99 <= 0.03 (absolute error of the output flow).

Usage

from fastforks import flow

model = flow.load("recover-cx")  # mode="auto": the preferred backend from recorded measurements
result = model("frame1.png", "frame2.png")
result.flow  # [2, H, W] float32 tensor: dx, dy in input pixels
result.backend  # what ran, e.g. Backend(mode='portable', device='cuda:0', dtype='float16')

Install with pip install fastforks (see https://github.com/kleinicke/fastforks). Other variants: recover-cx, recover-mn, recover-rn.

Kernels backend

flow.load("recover-cx", mode="kernels") runs our CUDA kernels (cudnn_hacc). It needs a prebuilt kernel for your GPU architecture (downloaded from https://huggingface.co/fastforks/kernels 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.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support

Collection including fastforks/recover