Update README.md

README.md

@@ -8,8 +8,8 @@ The model is designed to recognize characters from image inputs and can be used
 
 - **Parameter count**: 1.3M
 - **Type**: Convolutional Neural Network (CNN)
-- **Input**: Grayscale image of a single Armenian character (e.g., 28x28 or 64x64)
-- **Output**: One of the 76 classes:
+- **Input**: Grayscale image of a single Armenian character (64x64)
+- **Output**: One of the 78 classes:
   - 39 uppercase Armenian letters (Ա–Ֆ)
   - 39 lowercase Armenian letters (ա–ֆ)
   - Model's accuracy score is 0.95 (95%)