🕋🚨 Rock-Paper-Scissors Gesture Detection using TensorFlow

This project demonstrates a Computer Vision model that detects hand gestures: Rock, Paper, and Scissors using image classification. The model is trained using the official TensorFlow Rock-Paper-Scissors Dataset.

---

📁 Dataset

The dataset is provided by TensorFlow:

• Training Set: 840 images per class (rock, paper, scissors)
• Testing Set: 124 images per class

📦 Dataset Download Commands

!mkdir ./tmp

!wget --no-check-certificate \
  https://storage.googleapis.com/download.tensorflow.org/data/rps.zip \
  -O ./tmp/rps.zip

!wget --no-check-certificate \
  https://storage.googleapis.com/download.tensorflow.org/data/rps-test-set.zip \
  -O ./tmp/rps-test-set.zip

🗂️ Extract Dataset

import zipfile

def extract_file(src, dest):
    with zipfile.ZipFile(src, 'r') as zip:
        print(f'Extracting all the files from {src}...')
        zip.extractall(dest)
        print('Done!')

extract_file('./tmp/rps.zip', './data')
extract_file('./tmp/rps-test-set.zip', './data')

📊 Image Count

import os

def get_image_counts(parent_folder, dataset_name):
    print(f"total {dataset_name} rock images: {len(os.listdir(os.path.join(parent_folder, 'rock')))}")
    print(f"total {dataset_name} paper images: {len(os.listdir(os.path.join(parent_folder, 'paper')))}")
    print(f"total {dataset_name} scissors images: {len(os.listdir(os.path.join(parent_folder, 'scissors')))}")

get_image_counts('./data/rps', 'training')
get_image_counts('./data/rps-test-set', 'testing')

Gesture	Training Images	Testing Images
Rock	840	124
Paper	840	124
Scissors	840	124

---

🧠 Model Summary

• Implemented using TensorFlow & Keras

• Convolutional Neural Network (CNN)

• Categorical classification (3 classes)

• Trained using image data from the dataset

• Designed to predict whether a hand gesture is:

  • Rock 🪨
  • Paper 📄
  • Scissors ✂️

---

🖼️ Visualize Sample Images

from PIL import Image
import matplotlib.pyplot as plt
import numpy as np

for label in ['rock', 'paper', 'scissors']:
    im_folder = f'./data/rps/{label}'
    for im_name in os.listdir(im_folder)[:2]:
        img = Image.open(os.path.join(im_folder, im_name)).convert('RGB')
        plt.imshow(np.asarray(img))
        plt.title(f'Label: {label}')
        plt.axis('off')
        plt.show()

---

🏗️ CNN Model Architecture (Example)

model = keras.models.Sequential([
    keras.layers.Conv2D(32, (3,3), activation='relu', input_shape=(150, 150, 3)),
    keras.layers.MaxPooling2D(2,2),
    keras.layers.Conv2D(64, (3,3), activation='relu'),
    keras.layers.MaxPooling2D(2,2),
    keras.layers.Conv2D(128, (3,3), activation='relu'),
    keras.layers.MaxPooling2D(2,2),
    keras.layers.Flatten(),
    keras.layers.Dense(512, activation='relu'),
    keras.layers.Dense(3, activation='softmax')
])

---

📊 Model Evaluation

model.compile(loss='categorical_crossentropy',
              optimizer='adam',
              metrics=['accuracy'])

history = model.fit(train_generator,
                    epochs=10,
                    validation_data=validation_generator)

📈 Accuracy Visualization

plt.plot(history.history['accuracy'], label='accuracy')
plt.plot(history.history['val_accuracy'], label='val_accuracy')
plt.xlabel('Epoch')
plt.ylabel('Accuracy')
plt.legend()
plt.show()

---

🧪 Live Prediction (Optional)

Use OpenCV for real-time hand gesture prediction via webcam:

import cv2

cap = cv2.VideoCapture(0)
while True:
    ret, frame = cap.read()
    if not ret:
        break
    # Process frame here
    cv2.imshow('Rock-Paper-Scissors Detection', frame)
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break
cap.release()
cv2.destroyAllWindows()

---

✅ Project Checklist

• Download and extract Rock-Paper-Scissors dataset
• Display sample training images
• Count number of images per class
• Build and train a TensorFlow model
• Evaluate performance
• (Optional) Integrate with webcam

---

🚀 Getting Started

1. Open the notebook in Google Colab or your Jupyter environment.
2. Run all code cells step-by-step.
3. Extend with webcam support or mobile deployment.

---

📌 Notes

This was built as part of a Computer Vision assignment focused on gesture recognition using CNNs.

---

📑 References

• TensorFlow Dataset
• Keras Documentation
• OpenCV Documentation