YOLO12: Optimized for AMD ROCm
YOLO12 is a real-time object detection model that detects and localizes objects across 80 COCO categories in a single forward pass. This repository packages inference for object detection using ONNX Runtime, exported and validated for AMD ROCm so it runs efficiently on AMD GPUs, CPUs, and NPUs.
This is based on the implementation of YOLO12 found
here.
This repository contains configurations and scripts optimized for
AMD® ROCm™ platforms. You can use the
yolov12 AMD scripts to reproduce results or export with custom configurations. More details on model performance can be found
here.
Task Overview
Task: Object detection
Dataset: COCO val2017 (5,000 images, 80 categories)
Output metrics: mAP@0.5:0.95, mAP@0.5, mAP@0.75, Precision, Recall (AR@100), per-size mAP (small/medium/large)
Model variants: Default is m. Override with MODEL_SIZE=s/m/l/x.
AMD ROCm Optimization
This model export has been adapted and validated for AMD Instinct™ / Radeon™ GPUs running ROCm, as well as AMD CPUs and AMD Ryzen AI NPUs. Key points:
- Validated backends: ONNX Runtime across CPU (FP32), GPU (MIGraphX Execution Provider — FP32/FP16/BF16/INT8), and NPU (VitisAI Execution Provider, auto-quantized internally).
- No code changes required versus the upstream Ultralytics YOLO12 implementation — only environment/runtime configuration differs.
- First NPU run takes 5–10 minutes for model compilation; subsequent runs use the cached compiled model.
| Runtime | Precision | Backend | Hardware | Notes |
|---|
| ONNX Runtime | FP32 | CPU Execution Provider | AMD CPU | — |
| ONNX Runtime | FP32 / FP16 / BF16 / INT8 | MIGraphX Execution Provider | AMD Instinct™ / Radeon™ GPU (ROCm) | — |
| ONNX Runtime | Auto | VitisAI Execution Provider | AMD Ryzen AI NPU | Auto-quantized internally; first run compiles in 5–10 minutes |
Getting Started
For setup instructions, evaluation scripts, and custom configuration options, see the
yolov12 on GitHub.
Model Details
Model Type: Object detection (single-pass CNN detector)
Base Model: YOLO12 (Ultralytics), size m default
Model Stats:
- Input:
(1, 3, 640, 640) float32
- Output:
(1, 84, 8400) float32
- Precision tested: FP32 (CPU); FP32, FP16, BF16, INT8 (GPU); auto-quantized (NPU)
Accuracy Pipeline
COCO quality evaluation is fully implemented — make eval-fulldataset-<device> runs inference across all 5,000 COCO val2017 images and computes official COCO metrics via pycocotools.
Higher mAP means the model's predicted boxes and classes agree more closely with ground truth across the dataset — 1.0 would be perfect detection, 0.0 means no correct detections. In practice, values above ~0.5 for mAP@0.5:0.95 are considered strong for COCO-scale object detection.
Metrics Explained
| Metric | Description |
|---|
| mAP@0.5:0.95 | Primary COCO metric — mean AP averaged across IoU thresholds 0.5–0.95. The strictest, most holistic accuracy number; higher means boxes are both correctly classified and tightly localized across a range of overlap thresholds. |
| mAP@0.5 | AP at a single, looser IoU threshold of 0.5 (VOC-style) — a prediction only needs to overlap the ground-truth box by 50% to count as correct, so this is typically higher than mAP@0.5:0.95 and reflects "did it find the object" more than "how precisely." |
| mAP@0.75 | AP at a stricter IoU threshold of 0.75 — the predicted box must overlap ground truth by 75%, rewarding precise localization, not just correct detection. |
| Precision | Of all boxes the model predicted, what fraction were correct — high precision means few false positives. |
| Recall (AR@100) | Of all ground-truth objects, what fraction did the model find within its top 100 detections per image — high recall means few missed objects. |
| mAP-small/medium/large | mAP@0.5:0.95 broken down by object size — small objects are typically the hardest, exposing size-specific weaknesses a single aggregate score would hide. |
Accuracy Results
Full Dataset Evaluation (COCO val2017) — MODEL_SIZE=m:
| Device | Precision | mAP@0.5:0.95 | mAP@0.5 | mAP@0.75 | Precision | Recall |
|---|
| CPU | FP32 | 0.5141 | 0.6855 | 0.5568 | 0.4615 | 0.5623 |
| GPU | FP16 | 0.5137 | 0.6855 | 0.5555 | 0.4611 | 0.5616 |
| NPU | FP32 | 0.3164 | 0.4701 | 0.3458 | 0.2854 | 0.3582 |
Note: NPU quality may differ from CPU/GPU due to VitisAI's internal BF16 quantization. First NPU run takes 5–10 minutes for model compilation; subsequent runs use the cached compiled model.
Dig Deeper
Want to explore the full evaluation scripts, config options, and other AMD-optimized model examples?
The GitHub repository includes:
- Setup and prerequisites for ROCm environments
- Full COCO val2017 evaluation pipeline via
pycocotools
- Per-operator latency profiling scripts (including NPU AI Analyzer integration)
- Benchmarking and reproduction instructions across CPU, GPU, and NPU
