Object Detection
Keras
LiteRT
person-detection
search-and-rescue
drone
aerial
tinyml
int8
esp32
centernet
Instructions to use inviolus/voxOCR with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Keras
How to use inviolus/voxOCR with Keras:
# !pip install -U keras tensorflow huggingface_hub # Keras needs TensorFlow installed to read "hf://" paths, so the tensorflow backend is selected here; # "jax" and "torch" also work for computation once TensorFlow is installed. import os os.environ["KERAS_BACKEND"] = "tensorflow" import keras model = keras.saving.load_model("hf://inviolus/voxOCR") - Notebooks
- Google Colab
- Kaggle
Card: point to v3 on GitHub
Browse files
README.md
CHANGED
|
@@ -28,6 +28,11 @@ tile so a drone can flag places worth a closer look for a rescue team.
|
|
| 28 |
|
| 29 |
Code and training details: [ErdemWilkinson/voxOCR](https://github.com/ErdemWilkinson/voxOCR).
|
| 30 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 31 |
## Files
|
| 32 |
|
| 33 |
| File | What it is |
|
|
|
|
| 28 |
|
| 29 |
Code and training details: [ErdemWilkinson/voxOCR](https://github.com/ErdemWilkinson/voxOCR).
|
| 30 |
|
| 31 |
+
> **Newer version on GitHub (2026-11-01):** v3, trained 7 more hours from
|
| 32 |
+
> these weights, finds 83.9% of people at threshold 0.2 as int8 QAT, with
|
| 33 |
+
> 0.35 false alarms per tile (the files here, v2: 82.5% and 0.47). Scores:
|
| 34 |
+
> [voxOCR v3](https://github.com/ErdemWilkinson/voxOCR#v3-2026-11-01-7-more-hours-from-v2).
|
| 35 |
+
|
| 36 |
## Files
|
| 37 |
|
| 38 |
| File | What it is |
|