D-FINE (HGNetV2, N / S / M / L / X)
D-FINE (N, S, M, L, X; HGNetV2, COCO) from USTC: real-time end-to-end detector in the RT-DETR style, 300 queries, deformable attention and box edges refined as distributions, no NMS. Also DEIM (N, S, M, L, X): the same network with SiLU decoder MLPs, trained with dense one-to-one matching.
stable Fork of Peterande/D-FINE@956d170 · backends: portable, kernels
- Paper D-FINE: Redefine Regression Task in DETRs as Fine-grained Distribution Refinement ICLR 2025 · first released 17 Oct 2024
- Authors’ code Peterande/D-FINE (PyTorch), not the code forked here: The benchmarked original is the Hugging Face transformers port (DFineForObjectDetection), which loads the same weights; the repo itself was not run.
Paper 2024 Added in 0.1.0 Evaluated with 0.2.0
Use it
pip install fastforks Lean install with only what this model needs: pip install fastforks-core
from fastforks import detection
model = detection.load("dfine-n-coco") # kernels on a supported GPU, else portable PyTorch
result = model(image)
result.boxes # also: scores, labels
result.backend # what ran: mode, device, dtype, kernels fastforks bench dfine-n-coco Times the model on this machine. load(…, mode="portable") or mode="kernels" picks a backend; usage guide.
Speed and memory
Every backend on one GPU (RTX 5060 Ti 16 GB) and the same input, the authors’ code included. Kernels runs fp16. Portable runs fp16 on CUDA.
Input: synthetic, 800x1344 (the task’s standard size)
| Backend | dtype | Time per call (ms) | vs original fp32 | Peak memory allocated / reserved | Load (s) | First call (s) |
|---|---|---|---|---|---|---|
| Original fp32incl. CPU PIL resize + host-to-device copy of the HF-transformers pipeline (the official D-FINE .pth is not in the weights dir) | fp32 | 24.3 | – | 91 MiB/122 MiB | 0.2 | 0.35 |
| Original fp16torch.autocast(fp16) around the forward; incl. CPU PIL resize + host-to-device copy of the HF-transformers pipeline (the official D-FINE .pth is not in the weights dir); fp16 differs from the reference (box max diff 65.17 px) | fp16 | 29.1 | 0.83× | 92 MiB/126 MiB | 5.2 | 4.07 |
| Portable fp32 | fp32 | not measured | – | – | – | – |
| Portable | fp16 | 10.4 | 2.3× | 92 MiB/114 MiB | 1.1 | 0.80 |
| Kernels | fp16 | 2.48 | 9.8× | 94 MiB/118 MiB | 1.4 | 0.41 |
| Portable + torch.compile | 5.43 | 4.5× | 101 MiB/118 MiB | 27 (build) | – | |
| TensorRT fp16 | fp16 | 1.08 | 22× | 56 MiBenginenot comparable | 64 (build) | – |
Measured on RTX 5060 Ti 16 GB, torch 2.14.1+cu130, 2026-11-06 to 2026-11-16. First call, 3 warm-up, peak reset, median of 3 rounds x median of 5. Ratios divide two
times of this table (same GPU, same input). Original fp16 is the authors’ own precision option, else torch.autocast. torch.compile and TensorRT: build time instead of load time. TensorRT memory (engine, not comparable) is the GPU memory the engine took, read from the driver because TensorRT allocates outside PyTorch, so it can't be compared with the PyTorch peaks of the other rows.
Other devices
| Device | Backend | dtype | ms / call | Peak memory | Output vs reference | Input |
|---|---|---|---|---|---|---|
| Apple M1 Pro 16 GB MPS, torch 2.14.1, 2026-11-14 | Original | fp32 | 45.4 | 1,663 MiB process | – | 800x1344 |
| Portable | fp16 | 25.2 | 1,510 MiB process | matches | 800x1344 | |
| Portable | fp32 | 40.9 | 1,545 MiB process | matches | 800x1344 | |
| Kernels | fp16 | not implemented | – | 800x1344 | ||
| NVIDIA L4 CUDA, torch 2.14.1+cu130, 2026-11-07 | Portable | fp16 | 17.4 | 65 MiB/76 MiB | – | 800x1344 |
| Kernels | fp16 | 3.45 | 77 MiB/160 MiB | – | 800x1344 | |
| NVIDIA L40S CUDA, torch 2.14.1+cu130, 2026-11-07 | Portable | fp16 | 11.8 | 65 MiB/76 MiB | – | 800x1344 |
| Kernels | fp16 | 2.41 | 77 MiB/160 MiB | – | 800x1344 | |
| NVIDIA A10 CUDA, torch 2.14.1+cu130, 2026-11-07 | Portable | fp16 | 10.2 | 65 MiB/76 MiB | – | 800x1344 |
| Kernels | fp16 | 3.68 | 78 MiB/160 MiB | – | 800x1344 | |
| NVIDIA A100-SXM4-40GB CUDA, torch 2.14.1+cu130, 2026-11-07 | Portable | fp16 | 12.5 | 65 MiB/76 MiB | – | 800x1344 |
| Kernels | fp16 | 4.06 | 81 MiB/160 MiB | – | 800x1344 | |
| NVIDIA RTX PRO 6000 CUDA, torch 2.14.1+cu130, 2026-11-08 | Portable | fp16 | 4.38 | 88 MiB/116 MiB | – | 800x1344 |
| Kernels | fp16 | 1.89 | 126 MiB/200 MiB | – | 800x1344 | |
CUDA memory is torch’s peak allocated / reserved; Apple has no such counter, so the Mac rows show the process footprint (weights, Python and torch included). “Output vs reference” compares the device’s output with the stored golden output under the card’s tolerance; – = not checked on that device. Kernels “not implemented” = no kernels for this model on that backend yet: it runs portable there.
Licence and weights
Commercial use allowed
- Code
- Apache-2.0 LICENSELICENSE-DEIM
- Weights
- Apache-2.0 source
- Training data
-
- COCO 2017 (train, instance annotations) CC-BY-4.0
- ImageNet-1k (ILSVRC 2012 classification) noncommercial, non-commercial
Our reading of the licences, not legal advice.
Weights
dfine-nano-coco.safetensors 15.3 MB· hosted on Hugging Facefastforks downloads and checks these files on first use (fastforks download dfine-n-coco).
Accuracy
The test set through the authors’ code and through ours: same checkpoint, same protocol. The distance between the original’s fp32 and fp16 runs is what half precision alone costs; our runs should stay within it.
COCO val2017
every 10th image, box mAP
| Run | dtype | box mAP ↑ | Δ box mAP vs fp32 original | images | ms / image |
|---|---|---|---|---|---|
| Original fp32 | fp32 | 42.94 | reference | 500 | 19.1 |
| Original fp16 | fp16 | 42.91 | −0.03 | 500 | 23.6 |
| Portable fp32 | fp32 | 42.98 | +0.04 | 500 | 10.5 |
| Portable | fp16 | 42.91 | −0.03 | 500 | 10.2 |
| Kernels | fp16 | 42.97 | +0.03 | 500 | 2.67 |
pycocotools COCOeval of the top 100 detections per image. Original = the authors’ code, same checkpoint and protocol. Measured on original fp32, original fp16: RTX 5060 Ti, torch 2.14.1+cu130, 2026-11-11; portable fp32: RTX 5060 Ti, torch 2.14.1+cu130, fastforks 44640884, 2026-11-17; portable, kernels: RTX 5060 Ti, torch 2.14.1+cu130, fastforks 675d6e93, 2026-11-11.
Closeness to the original
Beyond the test-set scores: the raw output of ours against the original’s fp32 output on the same input, pixel by pixel. CPU, float32: max |diff| <= 0.002 vs the original. GPU, float16 (the default), portable and kernels: mean |diff| <= 0.2 and p99 <= 1.0 (absolute error of the output boxes).
| Check | Tolerance | Measured |
|---|---|---|
| CPU fp32, max |diff| | ≤ 0.002 | not recorded |
| GPU default dtypes, mean |diff| | ≤ 0.2 | 0.091 |
| GPU default dtypes, p99 |diff| | ≤ 1 | 0.32514 |
Recorded for dfine-n-coco on RTX 5060 Ti, torch 2.14.1+cu130, 2026-11-13, with these kernels: cudnn_haccfvm_conv
Details
D-FINE (Peng et al. 2025, USTC): real-time end-to-end detection like RT-DETR (HGNetV2 backbone, hybrid encoder, deformable-attention decoder over 300 selected queries, no NMS) whose decoder layers refine every box edge as a distribution over distances (fine-grained distribution refinement); boxes, scores and COCO classes at the input size. The portable `module()` is the whole network at 640x640 (class logits and normalised boxes).
- Input / output
- Input: one RGB image. Output: boxes, class labels, scores and instance masks.
- Input size
- any H x W >= 32, padded to a multiple of 32; tested up to 3000x4000 (498 MiB allocated)
- Options
score_threshold=0.050.0 to 1.0· changes the outputmax_detections=100>= 1· changes the outputcuda_graphs=true
- Outputs
boxes,scores,labels- Checkpoint
stem=[16,16],mid=[16,32,64,128],out=[64,256,512,1024],blocks=[1,1,2,1],layers=3,lab=true,levels=2,enc_dim=128,dec_dim=128,enc_ffn=512,dec_ffn=512,csp_blocks=2,expansion=0.34,dec_layers=3,points=[6,6]- CUDA kernels
cudnn_haccconvolutions with fp16 accumulation (cuDNN)fvm_convfused convolutions for image classifiers
- Apple (MPS)
- fp32, fp16
Limitations
- The original in the benchmarks and goldens is the Hugging Face transformers 5.19.0 port (DFineForObjectDetection, ustc-community weights) of the official D-FINE (Apache-2.0); the official repository code itself was not run.
- deim-* variants: DEIM (Huang et al. 2025), the D-FINE network with SiLU decoder MLPs, trained with dense one-to-one matching; weights and the original in the benchmarks are from the official DEIM repository (@09d35d53).
Full reference, generated from the repository
D-FINE (N, S, M, L, X; HGNetV2, COCO) from USTC: real-time end-to-end detector in the RT-DETR style, 300 queries, deformable attention and box edges refined as distributions, no NMS. Also DEIM (N, S, M, L, X): the same network with SiLU decoder MLPs, trained with dense one-to-one matching.
Commercial use: yes.
| variant | input | backend | time (ms) | peak memory (MiB, allocated / reserved) |
|---|---|---|---|---|
dfine-n-coco | synthetic (800x1344) | original (fp32) original portable kernels portable + torch.compile TensorRT fp16 | 24.3 29.1 10.4 2.48 5.43 1.08 | 91 / 122 92 / 126 92 / 114 94 / 118 101 / 118 104 / 128 |
dfine-s-coco | synthetic (800x1344) | original (fp32) original portable kernels portable + torch.compile TensorRT fp16 | 25.7 31.9 11.4 3.75 5.63 1.65 | 180 / 204 152 / 184 108 / 128 108 / 170 131 / 152 147 / 160 |
dfine-m-coco | synthetic (800x1344) | original (fp32) original portable kernels portable + torch.compile TensorRT fp16 | 30.8 38.2 13.7 5.07 6.79 2.49 | 231 / 264 209 / 256 129 / 142 129 / 184 149 / 166 204 / 232 |
dfine-l-coco | synthetic (800x1344) | original (fp32) original portable kernels portable + torch.compile TensorRT fp16 | 37.1 44.9 16.6 6.53 7.91 3.19 | 281 / 328 278 / 350 155 / 172 151 / 216 175 / 198 219 / 256 |
dfine-x-coco | synthetic (800x1344) | original (fp32) original portable kernels portable + torch.compile TensorRT fp16 | 44.9 54.6 18.6 9.59 11.0 5.64 | 452 / 488 476 / 536 230 / 258 231 / 298 245 / 284 342 / 374 |
deim-n-coco | synthetic (800x1344) | original portable kernels portable + torch.compile TensorRT fp16 | 24.6 10.4 2.46 5.44 1.10 | 93 / 124 92 / 114 94 / 118 101 / 118 104 / 128 |
deim-s-coco | synthetic (800x1344) | original portable kernels portable + torch.compile TensorRT fp16 | 26.4 11.3 3.73 5.63 1.68 | 154 / 180 108 / 128 108 / 170 131 / 152 163 / 176 |
deim-m-coco | synthetic (800x1344) | original portable kernels portable + torch.compile TensorRT fp16 | 31.4 13.8 5.08 6.73 2.54 | 210 / 250 129 / 142 129 / 184 149 / 166 204 / 232 |
deim-l-coco | synthetic (800x1344) | original portable kernels portable + torch.compile TensorRT fp16 | 36.1 16.3 6.50 7.99 3.39 | 264 / 322 155 / 172 151 / 216 175 / 198 229 / 266 |
deim-x-coco | synthetic (800x1344) | original portable kernels portable + torch.compile TensorRT fp16 | 41.3 18.7 9.53 11.0 5.10 | 462 / 502 230 / 258 231 / 298 245 / 284 352 / 394 |
Details
- Original code: https://github.com/Peterande/D-FINE at
956d1709314c - Code licence: Apache-2.0 (permissive); files:
LICENSE,LICENSE-DEIM - Weights licence: Apache-2.0
deim-n-coco: Apache-2.0 (assumed from code)deim-s-coco: Apache-2.0 (assumed from code)deim-m-coco: Apache-2.0 (assumed from code)deim-l-coco: Apache-2.0 (assumed from code)deim-x-coco: Apache-2.0 (assumed from code)
- Licence evidence: https://github.com/Peterande/D-FINE/blob/956d1709314c2c6a4df6f34de232054578a7449f/LICENSE
- Training data: COCO 2017 (train, instance annotations), ImageNet-1k (ILSVRC 2012 classification) (COCO 2017 train; the HGNetV2 backbones start from ImageNet-1k pretraining)
- Options:
score_threshold= 0.05 (0.0 to 1.0),max_detections= 100 (>= 1),cuda_graphs= True (speed only) - Kernel libraries:
cudnn_hacc,fvm_conv - Variants:
dfine-n-coco,dfine-s-coco,dfine-m-coco,dfine-l-coco,dfine-x-coco,deim-n-coco,deim-s-coco,deim-m-coco,deim-l-coco,deim-x-coco - Outputs: boxes, scores, labels; label_set:
coco-80 - Input size: any H x W >= 32, padded to a multiple of 32
- Tested up to (portable, fp16, RTX 5060 Ti 16 GB):
dfine-n-coco3000x4000 (0.5 GiB peak allocated, 0.6 reserved);dfine-s-coco3000x4000 (0.5 GiB peak allocated, 0.6 reserved);dfine-m-coco3000x4000 (0.5 GiB peak allocated, 0.6 reserved);dfine-l-coco3000x4000 (0.5 GiB peak allocated, 0.7 reserved);dfine-x-coco3000x4000 (0.6 GiB peak allocated, 0.7 reserved);deim-n-coco3000x4000 (0.5 GiB peak allocated, 0.5 reserved);deim-s-coco3000x4000 (0.5 GiB peak allocated, 0.5 reserved);deim-m-coco3000x4000 (0.5 GiB peak allocated, 0.5 reserved);deim-l-coco3000x4000 (0.5 GiB peak allocated, 0.6 reserved);deim-x-coco3000x4000 (0.6 GiB peak allocated, 0.6 reserved) dfine-n-coco800x1344, first call: portable + torch.compile 27 s (cold Inductor cache), TensorRT fp16 64 s (ONNX + engine).dfine-s-coco800x1344, first call: portable + torch.compile 31 s (cold Inductor cache), TensorRT fp16 51 s (ONNX + engine).dfine-m-coco800x1344, first call: portable + torch.compile 36 s (cold Inductor cache), TensorRT fp16 55 s (ONNX + engine).dfine-l-coco800x1344, first call: portable + torch.compile 46 s (cold Inductor cache), TensorRT fp16 59 s (ONNX + engine).dfine-x-coco800x1344, first call: portable + torch.compile 48 s (cold Inductor cache), TensorRT fp16 107 s (ONNX + engine).deim-n-coco800x1344, first call: portable + torch.compile 28 s (cold Inductor cache), TensorRT fp16 66 s (ONNX + engine).deim-s-coco800x1344, first call: portable + torch.compile 31 s (cold Inductor cache), TensorRT fp16 58 s (ONNX + engine).deim-m-coco800x1344, first call: portable + torch.compile 39 s (cold Inductor cache), TensorRT fp16 58 s (ONNX + engine).deim-l-coco800x1344, first call: portable + torch.compile 50 s (cold Inductor cache), TensorRT fp16 58 s (ONNX + engine).deim-x-coco800x1344, first call: portable + torch.compile 49 s (cold Inductor cache), TensorRT fp16 66 s (ONNX + engine).dfine-n-cocooriginal (fp32) 800x1344: incl. CPU PIL resize + host-to-device copy of the HF-transformers pipeline (the official D-FINE .pth is not in the weights dir)dfine-n-cocooriginal 800x1344: torch.autocast(fp16) around the forward; incl. CPU PIL resize + host-to-device copy of the HF-transformers pipeline (the official D-FINE .pth is not in the weights dir); fp16 differs from the reference (box max diff 65.17 px)dfine-s-cocooriginal (fp32) 800x1344: incl. CPU PIL resize + host-to-device copy of the HF-transformers pipeline (the official D-FINE .pth is not in the weights dir)dfine-s-cocooriginal 800x1344: torch.autocast(fp16) around the forward; incl. CPU PIL resize + host-to-device copy of the HF-transformers pipeline (the official D-FINE .pth is not in the weights dir)dfine-m-cocooriginal (fp32) 800x1344: incl. CPU PIL resize + host-to-device copy of the HF-transformers pipeline (the official D-FINE .pth is not in the weights dir)dfine-m-cocooriginal 800x1344: torch.autocast(fp16) around the forward; incl. CPU PIL resize + host-to-device copy of the HF-transformers pipeline (the official D-FINE .pth is not in the weights dir); fp16 differs from the reference (detections 5 vs reference 4)dfine-l-cocooriginal (fp32) 800x1344: incl. CPU PIL resize + host-to-device copy of the HF-transformers pipeline (the official D-FINE .pth is not in the weights dir)dfine-l-cocooriginal 800x1344: torch.autocast(fp16) around the forward; incl. CPU PIL resize + host-to-device copy of the HF-transformers pipeline (the official D-FINE .pth is not in the weights dir)dfine-x-cocooriginal (fp32) 800x1344: incl. CPU PIL resize + host-to-device copy of the HF-transformers pipeline (the official D-FINE .pth is not in the weights dir)dfine-x-cocooriginal 800x1344: torch.autocast(fp16) around the forward; incl. CPU PIL resize + host-to-device copy of the HF-transformers pipeline (the official D-FINE .pth is not in the weights dir)deim-n-cocooriginal 800x1344: torch.autocast(float16) around the call; official DEIM deploy model; incl. CPU PIL resize + host-to-device copy, RT-DETR postprocess (all 300)deim-s-cocooriginal 800x1344: torch.autocast(float16) around the call; official DEIM deploy model; incl. CPU PIL resize + host-to-device copy, RT-DETR postprocess (all 300)deim-m-cocooriginal 800x1344: torch.autocast(float16) around the call; official DEIM deploy model; incl. CPU PIL resize + host-to-device copy, RT-DETR postprocess (all 300)deim-l-cocooriginal 800x1344: torch.autocast(float16) around the call; official DEIM deploy model; incl. CPU PIL resize + host-to-device copy, RT-DETR postprocess (all 300)deim-x-cocooriginal 800x1344: torch.autocast(float16) around the call; official DEIM deploy model; incl. CPU PIL resize + host-to-device copy, RT-DETR postprocess (all 300)dfine-n-cocoload / first call: original 5.24 s / 4.072 s, portable 1.25 s / 0.853 s, kernels 1.21 s / 0.621 sdfine-s-cocoload / first call: original 2.99 s / 1.443 s, portable 1.24 s / 0.697 s, kernels 1.39 s / 0.358 sdfine-m-cocoload / first call: original 3.63 s / 1.546 s, portable 1.36 s / 0.702 s, kernels 1.49 s / 0.345 sdfine-l-cocoload / first call: original 3.68 s / 1.406 s, portable 1.72 s / 0.665 s, kernels 1.56 s / 0.322 sdfine-x-cocoload / first call: original 3.88 s / 1.452 s, portable 1.65 s / 0.642 s, kernels 1.83 s / 0.294 sdeim-n-cocoload / first call: original 2.39 s / 0.622 s, portable 1.61 s / 1.355 s, kernels 1.25 s / 0.583 sdeim-s-cocoload / first call: original 2.08 s / 0.534 s, portable 1.29 s / 0.687 s, kernels 1.43 s / 0.352 sdeim-m-cocoload / first call: original 2.24 s / 0.579 s, portable 1.38 s / 0.676 s, kernels 1.52 s / 0.346 sdeim-l-cocoload / first call: original 2.34 s / 0.516 s, portable 1.51 s / 0.665 s, kernels 1.61 s / 0.321 sdeim-x-cocoload / first call: original 2.63 s / 0.483 s, portable 1.72 s / 0.635 s, kernels 1.84 s / 0.286 s- Protocol, first call, 3 warm-up, peak reset, median of 3 rounds x median of 5
Measured
pycocotools COCOeval of the top 100 detections per image. Original = the original code named under Details, same checkpoint and protocol.
| variant | dataset | original fp32 | original fp16 | portable | portable fp32 | kernels | note |
|---|---|---|---|---|---|---|---|
dfine-n-coco | COCO val2017, every 10th image, box mAP | 42.94 | 42.91 | 42.91 | 42.98 | 42.97 | |
dfine-s-coco | COCO val2017, every 10th image, box mAP | 49.84 | 49.83 | 49.88 | 49.84 | 49.89 | |
dfine-m-coco | COCO val2017, every 10th image, box mAP | 51.92 | 51.95 | 52.03 | 51.92 | 52.00 | |
dfine-l-coco | COCO val2017, every 10th image, box mAP | 54.17 | 54.18 | 54.14 | 54.15 | 54.17 | |
dfine-x-coco | COCO val2017, every 10th image, box mAP | 56.40 | 56.38 | 56.30 | 56.35 | 56.32 | |
deim-n-coco | COCO val2017, every 10th image, box mAP | 44.38 | 44.37 | 44.40 | 44.36 | 44.38 | |
deim-s-coco | COCO val2017, every 10th image, box mAP | 49.46 | 49.45 | 49.50 | 49.36 | 49.52 | |
deim-m-coco | COCO val2017, every 10th image, box mAP | 52.38 | 52.43 | 52.43 | 52.38 | 52.43 | |
deim-l-coco | COCO val2017, every 10th image, box mAP | 54.67 | 54.65 | 54.67 | 54.67 | 54.67 | |
deim-x-coco | COCO val2017, every 10th image, box mAP | 56.87 | 56.91 | 56.86 | 56.88 | 56.90 |
Measured on portable: NVIDIA GeForce RTX 5060 Ti, torch 2.14.1+cu130, fastforks 675d6e93 / a981ff86, 2026-11-11 to 2026-11-12; portable32: NVIDIA GeForce RTX 5060 Ti, torch 2.14.1+cu130, fastforks 44640884, 2026-11-17; kernels: NVIDIA GeForce RTX 5060 Ti, torch 2.14.1+cu130, fastforks 675d6e93 / a981ff86, 2026-11-11 to 2026-11-12.
Parity
CPU, float32: max |diff| <= 0.002 vs the original. GPU, float16 (the default), portable and kernels: mean |diff| <= 0.2 and p99 <= 1.0 (absolute error of the output boxes).
| variant | kernels error vs original, mean / p99 | GPU, torch | kernels used |
|---|---|---|---|
dfine-n-coco | 0.091 / 0.32514 | RTX 5060 Ti, torch 2.14.1+cu130 | cudnn_hacc@cudnn92400, fvm_conv@f1d064e539e1:sm_120 |
dfine-s-coco | 0.05359 / 0.148 | RTX 5060 Ti, torch 2.14.1+cu130 | cudnn_hacc@cudnn92400, fvm_conv@f1d064e539e1:sm_120 |
dfine-m-coco | 0.06031 / 0.24838 | RTX 5060 Ti, torch 2.14.1+cu130 | cudnn_hacc@cudnn92400, fvm_conv@f1d064e539e1:sm_120 |
dfine-l-coco | 0.03714 / 0.09859 | RTX 5060 Ti, torch 2.14.1+cu130 | cudnn_hacc@cudnn92400, fvm_conv@f1d064e539e1:sm_120 |
dfine-x-coco | 0.03798 / 0.08887 | RTX 5060 Ti, torch 2.14.1+cu130 | cudnn_hacc@cudnn92400, fvm_conv@f1d064e539e1:sm_120 |
deim-n-coco | 0.07347 / 0.36832 | RTX 5060 Ti, torch 2.14.1+cu130 | cudnn_hacc@cudnn92400, fvm_conv@f1d064e539e1:sm_120 |
deim-s-coco | 0.04567 / 0.27886 | RTX 5060 Ti, torch 2.14.1+cu130 | cudnn_hacc@cudnn92400, fvm_conv@f1d064e539e1:sm_120 |
deim-m-coco | 0.04894 / 0.17919 | RTX 5060 Ti, torch 2.14.1+cu130 | cudnn_hacc@cudnn92400, fvm_conv@f1d064e539e1:sm_120 |
deim-l-coco | 0.03909 / 0.0618 | RTX 5060 Ti, torch 2.14.1+cu130 | cudnn_hacc@cudnn92400, fvm_conv@f1d064e539e1:sm_120 |
deim-x-coco | 0.05634 / 0.13909 | RTX 5060 Ti, torch 2.14.1+cu130 | cudnn_hacc@cudnn92400, fvm_conv@f1d064e539e1:sm_120 |
Limitations
- The original in the benchmarks and goldens is the Hugging Face transformers 5.19.0 port (DFineForObjectDetection, ustc-community weights) of the official D-FINE (Apache-2.0); the official repository code itself was not run.
- deim-* variants: DEIM (Huang et al. 2025), the D-FINE network with SiLU decoder MLPs, trained with dense one-to-one matching; weights and the original in the benchmarks are from the official DEIM repository (@09d35d53).
Example output
One fixed input through this checkpoint, the same for every model of the task (kernels · float16 · NVIDIA GeForce RTX 5060 Ti).
- person 0.93
- toothbrush 0.57
Input: Astronaut Eileen Collins, NASA (public domain).