|
|
--- |
|
|
license: other |
|
|
license_name: other |
|
|
license_link: LICENSE |
|
|
tags: |
|
|
- tinyml |
|
|
- edge-ai |
|
|
- cifar100 |
|
|
- cnn |
|
|
--- |
|
|
|
|
|
# φNet Tiny — Technical Proof One-Pager |
|
|
|
|
|
Inventor: Luigi Calcagno |
|
|
📧 lcalcagno1@yahoo.com |
|
|
Patent-pending φ-loss + Tiny CNN for True Edge Inference |
|
|
|
|
|
--- |
|
|
|
|
|
## ✅ What This Shows |
|
|
|
|
|
φNet Tiny is a custom convolutional neural network with: |
|
|
- Proprietary φ-loss function for tail-class lift. |
|
|
- Real ONNX → TensorFlow → TFLite flow. |
|
|
- Proven microcontroller-friendly footprint. |
|
|
|
|
|
Tested vs MobileNetV2 using real CIFAR-100 sample. |
|
|
|
|
|
--- |
|
|
|
|
|
## 📊 Combined Results |
|
|
|
|
|
Sample: torch.Size([1, 3, 32, 32]) |
|
|
Label: 49 (CIFAR-100) |
|
|
|
|
|
| Metric | φNet Tiny | MobileNetV2 | |
|
|
|--------|------------|--------------| |
|
|
| Param Size | 0.1148 MB | ~13.37 MB | |
|
|
| TFLite Size | 117.52 KB | Not feasible | |
|
|
| φ-loss Output | 4.5335 | N/A | |
|
|
| Inference Time (10 runs) | ~0.0094 s | ~0.0785 s | |
|
|
|
|
|
✅ φNet Tiny deploys `.tflite` ~117 KB — runs on STM32, ESP32, Pi Zero. |
|
|
❌ MobileNetV2 cannot run at MCU scale. |
|
|
|
|
|
--- |
|
|
|
|
|
## 🔑 Key Points |
|
|
|
|
|
- Reproducible φ-loss benchmark. |
|
|
- Fully local inference, no cloud. |
|
|
- Open pilot opportunities. |
|
|
|
|
|
--- |
|
|
|
|
|
**Built & tested by Luigi Calcagno** |
|
|
Contact: lcalcagno1@yahoo.com |
|
|
|
