https://github.com/akhdandann/rockpaperscissors-imageclassification

A Convolutional Neural Network (CNN)-based image classification project that identifies hand gestures from the classic Rock-Paper-Scissors game using TensorFlow and Keras. The model is trained on a labeled dataset and demonstrates image augmentation, model evaluation, and real-time prediction on uploaded images.
https://github.com/akhdandann/rockpaperscissors-imageclassification

cnn-classification image-classification keras python tensorflow

Last synced: 3 months ago
JSON representation

A Convolutional Neural Network (CNN)-based image classification project that identifies hand gestures from the classic Rock-Paper-Scissors game using TensorFlow and Keras. The model is trained on a labeled dataset and demonstrates image augmentation, model evaluation, and real-time prediction on uploaded images.

• Host: GitHub
• URL: https://github.com/akhdandann/rockpaperscissors-imageclassification
• Owner: akhdandann
• License: mit
• Created: 2025-04-08T08:22:29.000Z (about 1 year ago)
• Default Branch: main
• Last Pushed: 2025-05-14T03:59:13.000Z (about 1 year ago)
• Last Synced: 2025-06-17T21:45:38.253Z (about 1 year ago)
• Topics: cnn-classification, image-classification, keras, python, tensorflow
• Language: Jupyter Notebook
• Homepage:
• Size: 109 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# Rock-Paper-Scissors Image Classifier using CNN and Data Augmentation

This repository contains a TensorFlow + Keras implementation of a **Convolutional Neural Network (CNN)** model to classify images of hand gestures representing **Rock**, **Paper**, and **Scissors**.

## 📌 Description

This project is a practical image classification task where the model learns to distinguish between three categories of hand gestures: rock, paper, and scissors. The dataset is obtained from **Dicoding Academy's Rock-Paper-Scissors image set**, and the model uses **image augmentation** to improve generalization.

The model is trained on preprocessed images using `ImageDataGenerator` with transformations such as rotation, flipping, and zooming to simulate real-world variance.

The classification is performed using a **multi-class CNN** with 3 output nodes and `softmax` activation.

## 📁 Dataset

The dataset used is from Dicoding and publicly available:

- [Rock Paper Scissors Dataset](https://github.com/dicodingacademy/assets/releases/download/release/rockpaperscissors.zip)

It consists of labeled folders:

- `rock/`

- `paper/`

- `scissors/`

The code automatically downloads and extracts the dataset for training and validation.

## 🛠 Features

- Automatic dataset download and unzip

- Data augmentation using `ImageDataGenerator`

- Training-validation split with shuffling

- Convolutional Neural Network with 3 Conv layers

- Accuracy and loss graph visualization

- Image classification on user-uploaded image

## ⚙️ How It Works

1. **Download & Extract Data**

   - Uses `wget` and `zipfile` to fetch and unzip the dataset

2. **Data Exploration**

   - Verifies and prints folder structure

3. **Data Augmentation**

   - Applies rotation, zoom, flip, shear, etc. using `ImageDataGenerator`

4. **Model Architecture**

   - 3 convolutional layers + max pooling

   - Fully connected layer (512 neurons)

   - Output layer with 3 neurons + softmax

5. **Training**

   - Uses RMSprop optimizer and categorical crossentropy

   - Trains for 50 epochs with validation split

6. **Visualization**

   - Plots training & validation accuracy and loss

7. **Image Prediction**

   - Uploads image using `files.upload()` and classifies using the trained model

## 🚀 How to Run

1. Open the notebook in **Google Colab** or any Jupyter environment  

2. Install dependencies:

   ```python

   !pip install wget keras_preprocessing

   ```

3. Run the cells step by step:

   - Download and extract dataset  

   - Train the model  

   - Upload an image to test prediction  

## 📊 Output

- CNN model summary  

- Accuracy and loss charts  

- Prediction result of uploaded image (rock, paper, or scissors)  

## 📎 Dependencies

- TensorFlow  

- Keras Preprocessing  

- Numpy  

- Pandas  

- Matplotlib  

- wget  

- Google Colab (for `files.upload()`)

## 🧾 Note

This project was originally created as part of a **Dicoding certification test** around 5 years ago.  

While it may be slightly dated, the structure and techniques are still **very useful for those preparing for upcoming certification exams or learning basic CNNs with image data**.

Feel free to explore, learn, or build upon it!