DRCT (fastforks)

DRCT (Hsu et al., CVPRW 2024): image super-resolution with residual dense groups of Swin Transformer layers (16 x 16 window attention, shifted every second layer) whose inputs grow by 32 channels per layer (dense connections). drct-x4 is the classical x4 model (6 groups), drct-l-x4 the large one (12 groups); both bicubic, PSNR-trained.

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/ming053l/DRCT at c0a8b51e6924.

Task: restoration. Tier: experimental. Outputs: image. Backends: portable, kernels.

Input: one RGB image. Output: the restored (upscaled) image, model.scale x the input size, RGB in [0, 1]. tile= runs large images in overlapping tiles (as ComfyUI).

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
drct-l-x4@1 DRCT-L_X4.pth a99044c02756 MIT yes
drct-x4@1 DRCT_X4.pth fe104903aa8f MIT yes

Licence

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

Training data: ImageNet-1k (ILSVRC 2012 classification), DIV2K (NTIRE 2017 super-resolution, 1000 2K images), Flickr2K (2650 2K Flickr images, EDSR) (Paper / options/train/train_DRCT_SRx4_finetune_from_ImageNet_pretrain.yml: pretrained on ImageNet, fine-tuned on DF2K (DIV2K + Flickr2K), bicubic-downsampled. The README does not say which training option produced the released files.).

Licence evidence: https://github.com/ming053l/DRCT/blob/c0a8b51e69241cf7f8ae598053235ff498517935/LICENSE ; https://github.com/ming053l/DRCT/blob/c0a8b51e69241cf7f8ae598053235ff498517935/README.md (model zoo: https://drive.google.com/drive/folders/1QJHdSfo-0eFNb96i8qzMJAPw31u9qZ7U)

Input size

any H x W >= 32, padded to a multiple of 32 Tested up to (portable, fp16, RTX 5060 Ti 16 GB): drct-l-x4 1024x1024 (10.0 GiB peak allocated, 13.9 reserved); drct-x4 1024x1024 (10.0 GiB peak allocated, 13.9 reserved)

Speed and memory

Restoration, synthetic, 512x512

model time (ms): original portable kernels kernels speed-up: vs original vs portable peak memory (MiB, allocated / reserved): original portable kernels
drct-l-x4 14651 1607 1399 10.5x 1.1x 5574 / 6926 2692 / 3110 2564 / 2850
drct-x4 7492 830 718 10.4x 1.2x 3359 / 3944 2626 / 3048 2498 / 2766

Startup and first call

model load (s): original portable kernels first call (s): original portable kernels
drct-l-x4 2.09 0.79 1.84 15.38 1.70 1.45
drct-x4 1.88 0.39 1.32 8.66 1.02 0.77

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)
drct-l-x4 512x512 1607 ms 1210 ms deferred (build >5 min) 1399 ms 67.2 s – 2692 / 3110 2005 / 2410 2564 / 2850 –
drct-x4 512x512 830 ms 618 ms deferred (build >5 min) 718 ms 40.5 s – 2626 / 3048 1938 / 2348 2498 / 2766 –

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

DIV2K validation HR images and the official bicubic LR images (x4 or x2, the model's scale), one pass without tiles; PSNR and SSIM (Gaussian 11 x 11, sigma 1.5) on the Y channel (BT.601) of the output rounded to uint8, scale px border cropped, averaged over the images. Real-ESRGAN is trained for real-world degradations, not bicubic, so its PSNR is low: the numbers compare our modes with the original, not with PSNR-oriented upscalers. Original = the original code named under Details, same checkpoint and protocol.

variant dataset portable kernels note
drct-x4 DIV2K validation (every 2nd: 50 images), bicubic x4 LR -> HR, PSNR / SSIM (Y) 30.76 / 0.8388 30.76 / 0.8388
drct-l-x4 DIV2K validation (every 2nd: 50 images), bicubic x4 LR -> HR, PSNR / SSIM (Y) 31.00 / 0.8434 31.00 / 0.8434

Measured on NVIDIA GeForce RTX 5060 Ti, torch 2.14.1+cu130, fastforks 706c6828, 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| <= 2e-06 vs the original. GPU, float16 (the default), portable and kernels: mean |diff| <= 0.00071 and p99 <= 0.0027 (absolute error of the output image).

Usage

from fastforks import restoration

model = restoration.load("drct-l-x4")  # mode="auto": the preferred backend from recorded measurements
result = model("photo.png")  # tile=512 for large images
result.image  # [3, sH, sW] float32 RGB in [0, 1], s = model.scale
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: drct-l-x4, drct-x4.

Kernels backend

restoration.load("drct-l-x4", 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/drct