Deep Learning model to classify edible vs. poisonous mushrooms using Transfer Learning
The Mushroom Image Classifier is a deep learning–based computer vision project built to automatically identify whether a mushroom is edible or poisonous.
Using Transfer Learning with pretrained models like VGG16 and InceptionV3, this project leverages TensorFlow and Keras to deliver high accuracy and efficient classification, even on limited training data.
The main goal is to help in early identification of mushroom types using AI-assisted image recognition to reduce the risks of human error in field identification.
- ✅ Transfer Learning using TensorFlow/Keras
- 📷 Image preprocessing with augmentation
- 🧠 Model training, validation & testing pipeline
- 📊 Evaluation metrics (accuracy, loss, confusion matrix, F1-score)
- 💾 Model saving & real-time prediction support
- ☁️ Easily deployable on cloud or web apps
| Metric | Result |
|---|---|
| Accuracy | 96.8% |
| F1 Score | 0.97 |
| Model Used | InceptionV3 (fine-tuned) |
| Framework | TensorFlow / Keras |
Languages & Frameworks:
Python, TensorFlow, Keras, Matplotlib, Pandas, NumPy, scikit-learn
Techniques:
Transfer Learning, Convolutional Neural Networks, Data Augmentation, Model Fine-Tuning
The dataset contains thousands of labeled mushroom images belonging to two classes:
- 🍽️ Edible Mushrooms
- ☠️ Poisonous Mushrooms
It was preprocessed, resized, normalized, and split into Train / Validation / Test sets.