Create README.md

README.md

@@ -0,0 +1,96 @@
+---
+license: mit
+language:
+- en
+metrics:
+- accuracy
+- f1
+pipeline_tag: image-classification
+library_name: keras
+tags:
+- keras
+- classification
+- image_classification
+- CNN
+- image_recognition
+base_model:
+- timm/tf_efficientnetv2_m.in21k
+---
+
+# Indian Monuments CNN Model
+
+This model is a fine-tuned image classifier for recognizing major Indian monuments and architectural styles using EfficientNetV2-M as the base. It leverages transfer learning with Keras/TensorFlow, trained on the [danushkumarv/indian-monuments-image-dataset](https://www.kaggle.com/datasets/danushkumarv/indian-monuments-image-dataset) (24 classes).
+
+## Model Details
+
+- **Model Type:** Transfer Learning (EfficientNetV2-M)
+- **Task:** Image Classification (24 classes)
+- **Base Model:** `timm/tf_efficientnetv2_m.in21k`
+- **Dataset:** [danushkumarv/indian-monuments-image-dataset](https://www.kaggle.com/datasets/danushkumarv/indian-monuments-image-dataset)
+- **Framework:** Keras / TensorFlow
+
+## Results
+
+| Metric    | Value  |
+|-----------|--------|
+| Accuracy  | 0.921  |
+| F1 Score  | 0.918  |
+
+## Intended Uses
+
+- **Classification:** Predict image class for Indian monuments.
+- **Feature Extraction:** Use EfficientNet backbone for CV tasks.
+- **Education:** Integrate into apps/sites for learning about Indian heritage.
+
+## Limitations and Bias
+
+- **Out-of-Distribution:** Best for Indian monuments; may misclassify non-monument objects or unusual conditions.
+- **Class Imbalance:** Accuracy may favor classes with more samples.
+- **Fine-Grained Recognition:** Not for identifying sub-parts or rooms within monuments.
+
+## How to Use
+
+```python
+import keras
+from huggingface_hub import from_pretrained_keras
+
+repo_id = "koyelog/indian-monuments-cnn-model"
+model = from_pretrained_keras(repo_id)
+```
+
+```python
+import numpy as np
+from PIL import Image
+
+def preprocess_image(image_path, target_size=(224, 224)):
+    img = Image.open(image_path).convert('RGB')
+    img = img.resize(target_size)
+    img_array = np.asarray(img, dtype=np.float32)
+    img_array = img_array / 255.0
+    return np.expand_dims(img_array, axis=0)
+
+x = preprocess_image('path/to/your/monument.jpg')
+predictions = model.predict(x)
+predicted_class_index = np.argmax(predictions[0])
+
+# Define your class name mapping
+class_names = [
+    "Taj Mahal", "Red Fort", "Charminar", # ...add all class names
+]
+print(f"Predicted Monument: {class_names[predicted_class_index]}")
+```
+
+## Citation
+
+```
+@misc{koyelog_indian_monuments_cnn_model,
+  title={Indian Monuments CNN Model},
+  author={koyelog},
+  year={2025},
+  howpublished={\url{https://huggingface.co/koyelog/indian-monuments-cnn-model}}
+}
+```
+
+## Model Card Authors
+
+Model card generated by [koyelog](https://huggingface.co/koyelog)