WAFT (fastforks)
WAFT (Wang & Deng 2025), version a1: optical flow without a cost volume. A frozen Depth Anything V2 Small and a ResNet18 encoder give half-resolution features; each of 5 iterations warps the second image's features by the current flow and a small ViT-S (patch 8) refines the flow, which is convex-upsampled to full resolution. The authors' released weights; the network is ported from princeton-vl/WAFT. waft-a1 is the general checkpoint the authors recommend for downstream use (TartanAir, FlyingChairs, FlyingThings3D); -kitti and -sintel are their benchmark fine-tunes.
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/princeton-vl/WAFT at b152ff1cad1a.
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 |
|---|---|---|---|---|
waft-a1@1 |
waft-a1-tar-c-t.pth | 9f4b24f48b39 |
BSD-3-Clause (assumed from code) | yes |
waft-a1-kitti@1 |
waft-a1-tar-c-t-kitti.pth | b600f7821a52 |
BSD-3-Clause (assumed from code) | yes |
waft-a1-sintel@1 |
waft-a1-tar-c-t-sintel.pth | e3d95a6df6ec |
BSD-3-Clause (assumed from code) | 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.
Code licence: BSD-3-Clause; the authors do not state a separate weights licence.
Training data: TartanAir, FlyingChairs (Freiburg), Freiburg SceneFlow (FlyingThings3D, Driving, Monkaa) (TartanAir pretraining, FlyingChairs, then FlyingThings3D (part of SceneFlow), as the checkpoint name tar-c-t says; the Depth Anything V2 Small inside was trained by its authors on synthetic and pseudo-labelled real images).
Licence evidence: https://github.com/princeton-vl/WAFT/blob/b152ff1cad1af8c185ee7b141997c48ff3334c87/README.md#model-zoo
Input size
any H x W >= 112, resized or padded inside the model
Tested up to (portable, fp16, RTX 5060 Ti 16 GB): waft-a1 1080x1920 (3.5 GiB peak allocated, 4.0 reserved); waft-a1-kitti 1080x1920 (3.5 GiB peak allocated, 4.1 reserved); waft-a1-sintel 1080x1920 (3.5 GiB peak allocated, 4.0 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 |
|---|---|---|---|---|---|---|---|---|---|---|
waft-a1 |
1080x1920 | β | 1533 | 791 | 519 | 3.0x | 1.5x | 5537 / 5916 | 3553 / 4076 | 1851 / 2264 |
waft-a1-kitti |
1080x1920 | β | 1533 | 791 | β | β | β | 5537 / 5936 | 3553 / 4196 | β |
waft-a1-sintel |
1080x1920 | β | 1533 | 791 | β | β | β | 5537 / 5916 | 3553 / 4076 | β |
Startup and first call
| model | load (s): original | portable | kernels | first call (s): original | portable | kernels |
|---|---|---|---|---|---|---|
waft-a1 |
5.87 | 0.51 | 0.57 (375x1242) | 3.41 | 0.79 | 0.21 (375x1242) |
waft-a1-kitti |
4.06 | 0.54 | β | 2.87 | 0.79 | β |
waft-a1-sintel |
2.93 | 0.51 | β | 2.87 | 0.79 | β |
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) |
|---|---|---|---|---|---|---|---|---|---|---|---|
waft-a1 |
1080x1920 | 791 ms | 610 ms | deferred (build >5 min) | 519 ms | 84.7 s | β | 3553 / 4076 | 3479 / 3638 | 1851 / 2264 | β |
waft-a1-kitti |
1080x1920 | 791 ms | 610 ms | deferred (build >5 min) | β | 84.1 s | β | 3553 / 4196 | 3479 / 3758 | β | β |
waft-a1-sintel |
1080x1920 | 791 ms | 610 ms | deferred (build >5 min) | β | 84.7 s | β | 3553 / 4076 | 3479 / 3578 | β | β |
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 | note |
|---|---|---|---|---|---|
waft-a1 |
KITTI 2015 train, EPE / Fl-all % | 2.395 / 8.52 | 2.375 / 8.52 | 2.315 / 8.51 |
Measured on NVIDIA GeForce RTX 5060 Ti, torch 2.14.1+cu130, fastforks ec1d48ea, 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| <= 0.002 vs the original. GPU, float16 (the default), portable and kernels: mean |diff| <= 0.005 and p99 <= 0.02 (absolute error of the output flow).
Usage
from fastforks import flow
model = flow.load("waft-a1") # 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: waft-a1, waft-a1-kitti, waft-a1-sintel.
Kernels backend
flow.load("waft-a1", mode="kernels") runs our CUDA kernels (igev, cutlass_conv, cudnn_hacc, fastbb, dpt_fused). 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.