Image Classification
Transformers
Tibetan
tibetan
uchen
ume
script-classification
dinov3
fine-tuned
Eval Results (legacy)
Instructions to use openpecha/uchen-ume-classifier with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use openpecha/uchen-ume-classifier with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("image-classification", model="openpecha/uchen-ume-classifier") pipe("https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/hub/parrots.png")# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("openpecha/uchen-ume-classifier", dtype="auto") - Notebooks
- Google Colab
- Kaggle
Document confusion matrices in README
Browse files
README.md
CHANGED
|
@@ -103,6 +103,16 @@ Hub split: 9,110 train / 1,000 val / 851 test (work-stratified). Benchmark holdo
|
|
| 103 |
|
| 104 |
Run benchmark eval for `center_crop_all/` with `--preprocess center_crop_whole_page` to match training.
|
| 105 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 106 |
## Training data
|
| 107 |
|
| 108 |
| Class | Train | Validation | Test | Total |
|
|
|
|
| 103 |
|
| 104 |
Run benchmark eval for `center_crop_all/` with `--preprocess center_crop_whole_page` to match training.
|
| 105 |
|
| 106 |
+
### Test confusion matrices (851 pages)
|
| 107 |
+
|
| 108 |
+
| Variant | uchen→uchen | uchen→ume | ume→uchen | ume→ume |
|
| 109 |
+
|---------|------------:|----------:|----------:|--------:|
|
| 110 |
+
| **`center_crop_all/`** | 94 | 3 | 3 | 751 |
|
| 111 |
+
| **`without_preprocess/`** | 97 | 2 | 165 | 603 |
|
| 112 |
+
| `with_preprocess/` | 99 | 0 | 381 | 387 |
|
| 113 |
+
|
| 114 |
+
See `confusion_matrix.json` and `confusion_matrix.png` in each variant folder on the Hub.
|
| 115 |
+
|
| 116 |
## Training data
|
| 117 |
|
| 118 |
| Class | Train | Validation | Test | Total |
|