RF-DETR-Nano-OpenVINO

RF-DETR-Nano for Intel NPUs, GPUs and CPUs: rfdetr_graphA_fp16 and rfdetr_graphB_fp16 from litert-community/RF-DETR-Nano-LiteRT as ONNX (tf2onnx 1.17.0, opset 17), OpenVINO IR with fp32 weights, and precompiled Intel NPU blobs. tf2onnx 1.17.0 drops the bias of 5 FULLY_CONNECTED layers of rfdetr_graphA_fp16 and rfdetr_graphB_fp16; these files were converted with a one-line fix of that handler (convert/patch_tf2onnx.py in the GitHub repo), and their outputs match the .tflite. On the NPU of a Core Ultra 9 386H the blob runs in 18.54 ms (rfdetr_graphA_fp16) and 14.99 ms (rfdetr_graphB_fp16) per run, numpy input to numpy output.

Scripts, raw records, and the same table for all 52 graphs of LiteRT-Models: https://github.com/john-rocky/windows-zoo

Files

file	bytes	sha256
rfdetr_graphA_fp16.onnx	96,705,955	6bdbf0d3a0c943d484a404935d0b3119d8e52cc79003c6ee26efe3298a7723c7
openvino/rfdetr_graphA_fp16.xml	2,876,884	5f7f8119155043114216d8663fe86e9d40662e5756d56077affc98057d8ef79d
openvino/rfdetr_graphA_fp16.bin	93,922,448	06f1bcab729ad7152aec97d3070b715b5ac0949cf25eb3ea237aafe727538263
npu/rfdetr_graphA_fp16.blob	58,495,119	8af903d2fa7106339c3258488cff4823e1190ead5d12e10cc2d6a227e9b31f10
rfdetr_graphB_fp16.onnx	14,245,186	c60c3f9fee15d6cd5d3e1cdfdf280e0c0bcd42caa6f7fc802b2ebeb6e07163ae
openvino/rfdetr_graphB_fp16.xml	597,302	9260f0cd32551eed10b0da9a030c7671f03ecb9d6e2f251e598cc3291b8f64c3
openvino/rfdetr_graphB_fp16.bin	13,853,228	ea0db8800ceee469a00354443ad0a949cd08c18bf5a928b21650d78be415b8d7
npu/rfdetr_graphB_fp16.blob	12,058,767	2b637e1fcbe7f1bf65d6ef785ea92fb607ed6ce670ea7e220dcb3f738764c2cd

The NPU blobs were compiled with OpenVINO 2026.4.1, Intel NPU driver 32.0.100.4841 (OpenVINO reports NPU_DRIVER_VERSION 1004841, NPU_COMPILER_VERSION 524292). If import_model refuses a blob on another driver or OpenVINO version, compile the IR instead (the commented line in the snippet).

How it was measured

MSI Stealth 16 AI+ (Core Ultra 9 386H, RTX 5080 Laptop GPU at a 95 W power limit, Windows 11 build 26200.9711, on AC power), 2026-11-07. On the laptop, each row is one process: load, one run, 5 warmup runs, then the median of 30 timed runs (2 warmup + 10 timed runs when the first run takes more than 3 s), numpy input to numpy output, every runtime at its default settings, one process at a time, each started after the CPU, the RTX and the Intel GPU had gone idle (the thresholds are in METHOD.md of the GitHub repo). Output check: each output against the CPU output in fp32 on the same input (ONNX Runtime's CPU EP); PASS = cosine ≥ 0.99, no NaN or Inf, and top-1 agreement ≥ 0.99 for class scores. No Snapdragon rows: Qualcomm AI Hub measured the files before the bias fix (its numbers are in the GitHub table).

rfdetr_graphA_fp16

processor	runtime	file	ms	load s	output check
Intel AI Boost NPU	OpenVINO 2026.4.1	npu/ blob (import_model)	18.54	0.05	PASS (cos ≥ 0.9998)
Intel AI Boost NPU	ONNX Runtime 1.31.1 + OpenVINO EP 1.8.1	.onnx, compiled when the session loads	18.80	29.67	PASS (cos ≥ 0.9999)
Intel Graphics (iGPU)	ONNX Runtime 1.31.1 + OpenVINO EP 1.8.1	.onnx	16.10	2.04	PASS (cos ≥ 0.9999)
RTX 5080 Laptop (95 W limit)	ONNX Runtime 1.31.1, CUDA EP	fp16 copy of the .onnx (not in this repo)	14.25	1.36	PASS (cos ≥ 0.9999)
RTX 5080 Laptop (95 W limit)	ONNX Runtime 1.31.1, CUDA EP	.onnx	23.87	2.13	PASS (cos ≥ 0.9999)
Core Ultra 9 386H CPU	ONNX Runtime 1.31.1, CPU EP	.onnx	60.66	0.37	reference

The ONNX file against the source .tflite (LiteRT interpreter on the CPU, the same input): PASS, cosine ≥ 0.9999 for every output.

rfdetr_graphB_fp16

processor	runtime	file	ms	load s	output check
Intel AI Boost NPU	OpenVINO 2026.4.1	npu/ blob (import_model)	14.99	0.01	PASS (cos ≥ 0.9999)
Intel AI Boost NPU	ONNX Runtime 1.31.1 + OpenVINO EP 1.8.1	.onnx, compiled when the session loads	14.54	3.92	PASS (cos ≥ 0.9999)
Intel Graphics (iGPU)	ONNX Runtime 1.31.1 + OpenVINO EP 1.8.1	.onnx	3.90	1.01	PASS (cos ≥ 0.9999)
RTX 5080 Laptop (95 W limit)	ONNX Runtime 1.31.1, CUDA EP	fp16 copy of the .onnx (not in this repo)	3.44	1.22	PASS (cos ≥ 0.9999)
RTX 5080 Laptop (95 W limit)	ONNX Runtime 1.31.1, CUDA EP	.onnx	3.33	1.08	PASS (cos ≥ 0.9999)
Core Ultra 9 386H CPU	ONNX Runtime 1.31.1, CPU EP	.onnx	13.56	0.08	reference

The ONNX file against the source .tflite (LiteRT interpreter on the CPU, the same input): PASS, cosine ≥ 0.9999 for every output.

Run it on the Intel NPU

Download this repo, then run the snippets from its folder:

pip install huggingface_hub
hf download mlboydaisuke/RF-DETR-Nano-OpenVINO --local-dir RF-DETR-Nano-OpenVINO
cd RF-DETR-Nano-OpenVINO

OpenVINO (pip install openvino==2026.4.1):

import numpy as np
import openvino as ov

core = ov.Core()
if "NPU" not in core.available_devices:
    raise SystemExit(f"no Intel NPU visible to OpenVINO: {core.available_devices}")
with open("npu/rfdetr_graphA_fp16.blob", "rb") as f:
    compiled = core.import_model(f.read(), "NPU")  # precompiled: no compile at load
# or compile the IR on this machine: compiled = core.compile_model("openvino/rfdetr_graphA_fp16.xml", "NPU")
# replace the zeros with your preprocessed input (float32, the shapes printed by compiled.inputs)
feeds = {i.get_any_name(): np.zeros(list(i.shape), np.float32) for i in compiled.inputs}
outputs = compiled(feeds)
print({o.get_any_name(): outputs[o].shape for o in compiled.outputs})

ONNX Runtime with the OpenVINO EP (pip install onnxruntime==1.31.1 onnxruntime-ep-openvino==1.8.1):

import numpy as np
import onnxruntime as ort
import onnxruntime_ep_openvino as ov_ep

ort.register_execution_provider_library(ov_ep.get_ep_name(), ov_ep.get_library_path())
npu = [d for d in ort.get_ep_devices()
       if d.ep_name == ov_ep.get_ep_name() and d.ep_metadata.get("ov_device") == "NPU"]
if not npu:
    raise SystemExit("no Intel NPU visible to the OpenVINO EP")
so = ort.SessionOptions()
so.add_provider_for_devices(npu, {})
sess = ort.InferenceSession("rfdetr_graphA_fp16.onnx", sess_options=so)  # compiles for the NPU here
# replace the zeros with your preprocessed input (float32)
feeds = {i.name: np.zeros(i.shape, np.float32) for i in sess.get_inputs()}
outputs = sess.run(None, feeds)
print([o.shape for o in outputs])

The same lines run rfdetr_graphB_fp16 with its own file names.

Licence

Apache-2.0, the licence of litert-community/RF-DETR-Nano-LiteRT. The upstream model: Roboflow/rf-detr-nano.