Upload folder using huggingface_hub

README.md

@@ -2,8 +2,6 @@
 license: mit
 base_model:
 - google/efficientnet-b0
-datasets:
-- docling-project/HF-CC-v0-00001-00010-images-filtered-new-class
 tags:
 - image-classification
 - document-analysis
@@ -13,7 +11,7 @@ tags:
 
 # EfficientNet-B0 Document Figure Classifier v2.5
 
-This is an image classification model based on **Google EfficientNet-B0**, fine-tuned on a subset of the [subset of HuggingFace/finepdfs](https://huggingface.co/datasets/docling-project/HF-CC-v0-00001-00010-images-filtered-new-class) to classify document figures into one of the following 26 categories:
+This is an image classification model based on **Google EfficientNet-B0**, fine-tuned on a subset of the subset of HuggingFace/finepdfs to classify document figures into one of the following 26 categories:
 
 1. **logo**
 2. **photograph**
@@ -59,34 +57,34 @@ The model was evaluated on a held-out test set from the finepdfs dataset with th
 
 ### Per-Label Performance
 
-| Label | Precision | Recall |
-|-------|-----------|--------|
-| **logo** | 0.92807 | 0.91816 |
-| **photograph** | 0.90966 | 0.96029 |
-| **icon** | 0.83605 | 0.82678 |
-| **engineering_drawing** | 0.71689 | 0.81172 |
-| **line_chart** | 0.73055 | 0.92117 |
-| **bar_chart** | 0.88599 | 0.92720 |
-| **other** | 0.41893 | 0.38213 |
-| **table** | 0.98636 | 0.96765 |
-| **flow_chart** | 0.75926 | 0.82425 |
-| **screenshot_from_computer** | 0.85952 | 0.71980 |
-| **signature** | 0.89020 | 0.85971 |
-| **screenshot_from_manual** | 0.48559 | 0.34543 |
-| **geographical_map** | 0.86780 | 0.85219 |
-| **pie_chart** | 0.96880 | 0.94220 |
-| **page_thumbnail** | 0.52008 | 0.35188 |
-| **stamp** | 0.71269 | 0.41794 |
-| **music** | 0.48037 | 0.57778 |
-| **calendar** | 0.52880 | 0.28775 |
-| **qr_code** | 0.95694 | 0.93240 |
-| **bar_code** | 0.34244 | 0.84305 |
-| **full_page_image** | 0.40323 | 0.65789 |
-| **scatter_plot** | 0.66848 | 0.67213 |
-| **chemistry_structure** | 0.72781 | 0.65426 |
-| **topographical_map** | 0.83333 | 0.38462 |
-| **crossword_puzzle** | 0.57143 | 0.21622 |
-| **box_plot** | 0.85714 | 0.64286 |
+| Label | Precision (v2.5) | Recall (v2.5) | Precision (v2.0) | Recall (v2.0) |
+|-------|------------------|---------------|------------------|---------------|
+| **logo** | 0.92807 | 0.91816 | 0.88317 | 0.88728 |
+| **photograph** | 0.90966 | 0.96029 | 0.88169 | 0.93359 |
+| **icon** | 0.83605 | 0.82678 | 0.79281 | 0.72133 |
+| **engineering_drawing** | 0.71689 | 0.81172 | 0.58795 | 0.71555 |
+| **line_chart** | 0.73055 | 0.92117 | 0.75865 | 0.84576 |
+| **bar_chart** | 0.88599 | 0.92720 | 0.72624 | 0.93883 |
+| **other** | 0.41893 | 0.38213 | 0.28239 | 0.37312 |
+| **table** | 0.98636 | 0.96765 | 0.97950 | 0.95250 |
+| **flow_chart** | 0.75926 | 0.82425 | 0.61527 | 0.81518 |
+| **screenshot_from_computer** | 0.85952 | 0.71980 | 0.80510 | 0.65844 |
+| **signature** | 0.89020 | 0.85971 | 0.91852 | 0.80914 |
+| **screenshot_from_manual** | 0.48559 | 0.34543 | 0.34748 | 0.20662 |
+| **geographical_map** | 0.86780 | 0.85219 | 0.82959 | 0.80720 |
+| **pie_chart** | 0.96880 | 0.94220 | 0.89903 | 0.93931 |
+| **page_thumbnail** | 0.52008 | 0.35188 | 0.40194 | 0.21475 |
+| **stamp** | 0.71269 | 0.41794 | 0.63492 | 0.26258 |
+| **music** | 0.48037 | 0.57778 | 0.76955 | 0.51944 |
+| **calendar** | 0.52880 | 0.28775 | 0.51176 | 0.24786 |
+| **qr_code** | 0.95694 | 0.93240 | 0.97500 | 0.90909 |
+| **bar_code** | 0.34244 | 0.84305 | 0.12087 | 0.82063 |
+| **full_page_image** | 0.40323 | 0.65789 | 0.43750 | 0.28116 |
+| **scatter_plot** | 0.66848 | 0.67213 | 0.60386 | 0.68306 |
+| **chemistry_structure** | 0.72781 | 0.65426 | 0.77444 | 0.54787 |
+| **topographical_map** | 0.83333 | 0.38462 | 0.68750 | 0.28205 |
+| **crossword_puzzle** | 0.57143 | 0.21622 | 0.80000 | 0.21622 |
+| **box_plot** | 0.85714 | 0.64286 | 1.00000 | 0.07143 |
 
 
 ## How to use - Transformers
@@ -238,7 +236,7 @@ for item in ort_session.run(None, {'input': onnx_inputs}):
 
 ## Training Data
 
-This model was trained on a subset of the [subset of HuggingFace/finepdfs](https://huggingface.co/datasets/docling-project/HF-CC-v0-00001-00010-images-filtered-new-class), a large-scale dataset for document understanding tasks.
+This model was trained on a subset of the subset of HuggingFace/finepdfs, a large-scale dataset for document understanding tasks.
 
 
 ## Citation