Add model card for rfdetr-medium-fp16-coreml
Browse files
README.md
ADDED
|
@@ -0,0 +1,43 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
tags:
|
| 4 |
+
- coreml
|
| 5 |
+
- object-detection
|
| 6 |
+
- real-time
|
| 7 |
+
- rfdetr
|
| 8 |
+
- computer-vision
|
| 9 |
+
---
|
| 10 |
+
|
| 11 |
+
# Rfdetr Medium Fp16 Coreml
|
| 12 |
+
|
| 13 |
+
RF-DETR Medium model exported to CoreML format (FP16) for Apple devices. Best accuracy with 576×576 resolution.
|
| 14 |
+
|
| 15 |
+
## Model Details
|
| 16 |
+
|
| 17 |
+
- **Model Type**: Object Detection
|
| 18 |
+
- **Framework**: CoreML
|
| 19 |
+
- **Precision**: FP16
|
| 20 |
+
- **Input Resolution**: 576×576
|
| 21 |
+
- **Performance**: ~24 FPS on iPhone 13 Mini
|
| 22 |
+
- **Inference Time**: 41.85 ms
|
| 23 |
+
|
| 24 |
+
## Usage
|
| 25 |
+
|
| 26 |
+
This model can be used for real-time object detection on Apple devices (iOS, macOS, watchOS, tvOS).
|
| 27 |
+
|
| 28 |
+
|
| 29 |
+
## Custom conversion to CoreML
|
| 30 |
+
|
| 31 |
+
Check the repo https://github.com/pzoltowski/rf-detr-coreml
|
| 32 |
+
|
| 33 |
+
## Performance
|
| 34 |
+
|
| 35 |
+
| Model | Resolution | Inference Time (ms) | FPS |
|
| 36 |
+
|-------|------------|---------------------|-----|
|
| 37 |
+
| RF-DETR-Nano | 384×384 | 17.96 | ~56 |
|
| 38 |
+
| RF-DETR-Small | 512×512 | 27.07 | ~37 |
|
| 39 |
+
| RF-DETR-Medium | 576×576 | 41.85 | ~24 |
|
| 40 |
+
|
| 41 |
+
*Performance measured on iPhone 13 Mini using Xcode performance tools with FP16 precision.*
|
| 42 |
+
|
| 43 |
+
```
|