🌸🌻🌹 Multi-Class Image Classification 🌼🌷

💥 Computer Vision Project 💥

🌼 Flower Recognition Project 🌼

This project implements a comprehensive workflow for flower recognition using a combination of traditional and deep learning techniques.

📂 Dataset

The flower recognition dataset (tf_flowers) is loaded from TensorFlow Datasets. It contains images of five types of flowers: 🌼 daisies, 🌷 tulips, 🌻 sunflowers, 🌾 dandelions, and 🌹 roses.

🛠️ Steps

1️⃣ Load Flower Recognition Dataset

We load and split the dataset into training and testing sets.

2️⃣ Count Classes and Images

We count the number of classes and examples in the training and test sets, and plot the class distribution.

3️⃣ Data Augmentation and Preprocessing

Images are augmented (rotated, flipped, resized) and preprocessed (normalized, converted to grayscale, histogram equalization, and denoised).

4️⃣ Feature Extraction

We extract three types of features from each image:

• HOG Features: Describes the structure and appearance of the images.
• LBP Features: Captures texture information.
• CNN Features: Extracted using a pre-trained VGG16 model.

5️⃣ Dimensionality Reduction

Features are reduced using:

• PCA (Principal Component Analysis)
• LDA (Linear Discriminant Analysis)
• ICA (Independent Component Analysis)

6️⃣ Classification

We train RandomForest and SVM classifiers on the reduced feature sets (PCA, LDA, ICA) and evaluate their performance using accuracy, precision, recall, and F1-score.

📊 Results Visualization

Results are visualized using bar plots to compare the performance of different dimensionality reduction techniques and SVM kernels.

📋 Requirements

• 🐍 Python 3.x
• 🤖 TensorFlow
• 📚 TensorFlow Datasets
• 📷 OpenCV
• 🖼️ scikit-image
• 📈 scikit-learn
• 📊 Matplotlib
• 🔢 NumPy

🚀 How to Run

1. 📥 Clone this repository.
2. 📦 Install the required libraries.
3. ▶️ Run the main script to execute the workflow.

🏆 Conclusion

This project demonstrates an effective pipeline for flower recognition using a combination of traditional image processing techniques and deep learning models. The results indicate that the choice of dimensionality reduction technique and classifier significantly impacts the performance of the model.

📄 License

This project is licensed under the MIT License.

Made with 💖 by Hamza Sajjad.