Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -1,3 +1,31 @@
|
|
| 1 |
-
---
|
| 2 |
license: mit
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
license: mit
|
| 3 |
+
tags:
|
| 4 |
+
- onnx
|
| 5 |
+
- time-series
|
| 6 |
+
- imu
|
| 7 |
+
- navigation
|
| 8 |
+
- dead-reckoning
|
| 9 |
---
|
| 10 |
+
|
| 11 |
+
# COAST-VNet-1 (Tiny)
|
| 12 |
+
|
| 13 |
+
Edge-native, frequency-decoupled AI model for two-wheeler dead reckoning in GNSS-denied environments (urban canyons, tunnels, underpasses).
|
| 14 |
+
|
| 15 |
+
## Model Details
|
| 16 |
+
- **Architecture:** Frequency-Decoupled CNN-GRU
|
| 17 |
+
- **Parameters:** 96,086
|
| 18 |
+
- **Size:** 392 KB
|
| 19 |
+
- **Format:** ONNX (Optimized for Mobile/Edge CPU)
|
| 20 |
+
- **Input:** (20, 6) tensor — 2.0s window @ 10 Hz (ax, ay, az, gx, gy, gz)
|
| 21 |
+
- **Output:** (6) — Forward speed, Yaw rate, Roll/Pitch residuals, and Log-Variance (Uncertainty).
|
| 22 |
+
|
| 23 |
+
## Training Data
|
| 24 |
+
Trained on the IO-VNBD Dataset (23 drives). Ground truth labels were sourced exclusively from the vehicle CAN-bus, completely bypassing noisy phone GPS signals to ensure the model learns true kinematics.
|
| 25 |
+
|
| 26 |
+
## Use Case
|
| 27 |
+
Designed to maintain navigation continuity for ride-hailing and emergency response fleets. The model mathematically separates low-frequency vehicle motion (lean/acceleration) from high-frequency road and engine vibration.
|
| 28 |
+
|
| 29 |
+
## Performance
|
| 30 |
+
- **Inference Speed:** 8.3 microseconds per step (19,682 Hz on standard CPU).
|
| 31 |
+
- **Accuracy Gain:** 2.02x lower median position error compared to naive dead reckoning over 43 real-world GPS outages.
|