Downloads · 30 days
6
43% of all-time downloads
rf-detr-large
AXERA-TECH/rf-detr-large
rf-detr-large is a machine learning model from AXERA-TECH. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
This version of RF-DETR has been converted to run on the Axera NPU using u8 (INT8) quantization.
All-time downloads
14
Public
Repo size
309 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.onnx243 MB · 79%
At a glance
- Access
- Public
- Created
- Aug 21, 2026
- Updated
- Aug 23, 2026
- SHA
- 6a71d9f1
Model card
From the Hugging Face model README
RF-DETR Large
This version of RF-DETR has been converted to run on the Axera NPU using u8 (INT8) quantization.
Converted with Pulsar2 version 7.0.
原模型
git clone https://huggingface.co/Roboflow/rf-detr-large
切分模型
| 模型 | 作用 | 输入 | 输出 |
|---|---|---|---|
models/rf-detr-large_704_sim.onnx | 切分源模型／完整 FP32 基准模型 | pixel_values，FP32 NCHW，704x704 | logits、pred_boxes |
models/rf-detr-large_704_b2_pulsar2.onnx | pulsar2 编译输入图 | pixel_values | logits、bbox_features、reference_boxes |
models/rf-detr-large_704_b2_post.onnx | Host bbox 后处理图 | bbox_features、reference_boxes | pred_boxes |
logits 为分类输出；pred_boxes 为归一化的 cxcywh 框。B2 后处理图仅负责由中间 bbox 特征和参考框生成 pred_boxes。
脚本
| 脚本 | 运行位置 | 作用 |
|---|---|---|
src/tool/export_rfdetr_onnx_hf.py | 开发机 | 从 Hugging Face 默认缓存加载原始 large 模型，按配置读取 704x704 输入并使用 onnxsim 优化固定 shape 的 ONNX。 |
src/tool/b2_cut.py | 开发机 | 从完整 ONNX 提取 pulsar2 输入图和 Host 后处理图，并按中间张量 shape 识别 large 的切点。 |
src/infer/b2_infer_onnx.py | 开发机 | 使用 onnxruntime 运行 FP32 ONNX，保存 raw_outputs.npz 与性能数据。 |
src/infer/b2_infer_axmodel.py | AX650 | 使用 axengine 运行 .axmodel；指定 --post-model 时调用 Host 后处理图。 |
src/tool/evaluate_rfdetr.py | 开发机 | 将 raw_outputs.npz 转为 COCO 预测、计算 mAP，并比较两份 metrics.json。 |
编译配置为 pulsar2_config.json，校准集使用 datasets/coco2017val_64_calibration_dataset.tar。
结果展示
rf-detr-large.axmodel板端运行结果
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.490
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.681
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.528
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.272
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.544
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.697
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.367
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.597
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.648
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.404
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.716
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.873
=== 正式 COCO mAP 结果 ===
原始输出: results/ax/raw_outputs.npz
图数=5000 预测数=1499905
AP@[.5:.95]=0.4896 [email protected]=0.6810 [email protected]=0.5280
AP_small=0.2719 AP_medium=0.5441 AP_large=0.6973
rf-detr-large.axmodel推理速度评估
| 模型 | CMM(MB) | 延迟（ms,Avg） | FPS | 测试条件 |
|---|---|---|---|---|
| rf-detr-large.axmodel | 62.3 | 28.929 | 34 | AX650, 输入704×704, batch=1 |
rf-detr-large.axmodel对比ONNX运行结果
python src/tool/evaluate_rfdetr.py --compare-metrics \
results/onnx/metrics.json results/ax/metrics.json \
--primary -metric 'AP@[0.5:0.95]'
量化精度正式对比
评测集: datasets/annotations_val2017/annotations/instances_val2017.json
图片数: 5000
metric | baseline | quantized | drop
-----------------|-----------|-----------|---------
[email protected] | 0.709629 | 0.680995 | 0.028634
[email protected] | 0.564776 | 0.528006 | 0.03677
AP@[0.5:0.95] | 0.522527 | 0.489605 | 0.032923
AP_large | 0.723467 | 0.697302 | 0.026165
AP_medium | 0.575634 | 0.544072 | 0.031562
AP_small | 0.312324 | 0.271881 | 0.040443
AR@1 | 0.382697 | 0.366594 | 0.016103
AR@10 | 0.621168 | 0.596892 | 0.024276
AR@100 | 0.672812 | 0.648423 | 0.024389
AR_large | 0.87384 | 0.872696 | 0.001144
AR_medium | 0.73998 | 0.715606 | 0.024374
AR_small | 0.447414 | 0.403977 | 0.043437
prediction_count | 1499568.0 | 1499905.0 | -337.0
基本流程
# 1. 下载模型
git clone https://huggingface.co/AXERA-TECH/rf-detr-large
cd rf-detr-large
# 2. 从 Hugging Face 下载原始模型、导出并优化 ONNX
python src/tool/export_rfdetr_onnx_hf.py
# 3. 切分 ONNX
python /b2_cut.py \
--input models/rf-detr-large_704_sim.onnx \
--pulsar2-output models/rf-detr-large_704_b2_pulsar2.onnx \
--post-output models/rf-detr-large_704_b2_post.onnx
# 4. 用 pulsar2 编译 rf-detr-large_704_b2_pulsar2.onnx
# 配置：pulsar2_config.json
pulsar2 build --target_hardware AX650 --input rf-detr-large_704_b2_pulsar2.onnx --output_dir output/ --config pulsar2_config.json
# 5. 板端推理（脚本、模型、后处理 ONNX 与图片目录位于同级目录）
python3 b2_infer_axmodel.py \
--axmodel rf-detr-large.axmodel --post-model rf-detr-large_704_b2_post.onnx \
--img-dir coco2017val_5000 --result-dir results/ax
# 6. 开发机 COCO mAP 评测
python src/tool/evaluate_rfdetr.py/ \
--input results/onnx/raw_outputs.npz \
--ann datasets/annotations_val2017/annotations/instances_val2017.json \
--result-dir results/onnx
# 7. 对比 FP32 与量化后的 metrics.json
python src/tool/evaluate_rfdetr.py \
--compare-metrics \
results/onnx/metrics.json results/ax/metrics.json \
--primary-metric 'AP@[0.5:0.95]'
- Created
- Aug 21, 2026
- Updated
- Aug 23, 2026