Adhithpasu/DigitRecognition

Digit Recognition Model

Model Summary

A convolutional neural network (CNN) trained to classify handwritten digits (0–9) from image input. Trained on the MNIST dataset, this model serves as a foundational computer vision project demonstrating image classification with deep learning.

---

Model Details

• Developed by: Chandrasekar Adhithya Pasumarthi (@Adhithpasu)
• Affiliation: Frisco ISD, TX | AI Club Leader | Class of 2027
• Model type: Convolutional Neural Network (CNN)
• Framework: TensorFlow / Keras
• License: Apache 2.0
• Related work: Part of a broader ML/CV portfolio including research on Vision Transformers vs CNNs — JCSTS Vol. 8(2), January 2026

---

Intended Uses

Direct use:

• Handwritten digit classification (0–9)
• Educational demonstrations of CNNs and image classification pipelines
• Baseline model for comparing against more advanced architectures (ViT, ResNet, etc.)

Out-of-scope use:

• Multi-character or multi-digit recognition (e.g., full number strings)
• Non-MNIST-style digit distributions without fine-tuning

---

Training Data

Trained on the MNIST dataset — 60,000 training images and 10,000 test images of 28×28 grayscale handwritten digits.

---

Evaluation

Metric	Value
Accuracy	TBD
Loss	TBD

(Fill in with your actual test set results)

---

How to Use

import tensorflow as tf
import numpy as np
from PIL import Image

# Load model
model = tf.keras.models.load_model("digit_recognition_model")

# Load and preprocess a 28x28 grayscale image
img = Image.open("digit.png").convert("L").resize((28, 28))
img_array = np.array(img) / 255.0
img_array = img_array.reshape(1, 28, 28, 1)

# Predict
prediction = model.predict(img_array)
print(f"Predicted digit: {np.argmax(prediction)}")

---

Model Architecture

Input (28x28x1)
→ Conv2D(32, 3x3, relu) → MaxPooling2D
→ Conv2D(64, 3x3, relu) → MaxPooling2D
→ Flatten
→ Dense(128, relu) → Dropout(0.5)
→ Dense(10, softmax)

(Update to match your actual architecture)

---

Limitations & Bias

• Performs best on MNIST-style centered, normalized digits
• May degrade on real-world handwriting without preprocessing or fine-tuning
• Limited to single digit classification

---

Citation

@misc{pasumarthi2026digitrecognition,
  author    = {Chandrasekar Adhithya Pasumarthi},
  title     = {Digit Recognition Model},
  year      = {2026},
  publisher = {Hugging Face},
  url       = {https://huggingface.co/Chandrasekar123/DigitRecognition}
}

---

Contact

• GitHub: @Adhithpasu