An open API service indexing awesome lists of open source software.

https://github.com/rickydoan/deep-learning-cnn-calltech101-pretrained-transfer-learning


https://github.com/rickydoan/deep-learning-cnn-calltech101-pretrained-transfer-learning

Last synced: 11 months ago
JSON representation

Awesome Lists containing this project

README

          

# 🖼️ Caltech101 Image Classification with CNNs and Transfer Learning
📋 Overview
* This project involves image classification on the Caltech101 dataset, exploring both custom CNNs and pre-trained models to classify 102 object categories effectively.

# 🛠 Key Methods and Workflow
#Data Preparation
* Images resized to 128x128 pixels.
* Data augmentation (flips, rotations) applied to improve generalization.

# Custom CNN
* Built a CNN with convolutional, pooling, and fully connected layers.
* Trained for 5 epochs, achieving 45% accuracy.
* Transfer Learning

# Fine-tuned ResNet18 and EfficientNet-B0:
* ResNet18: 63% accuracy.
* EfficientNet-B0: 85% accuracy.

# Evaluation
* Compared accuracy and loss across models.
* Visualized predictions on random images with both true and predicted labels.
# 🔑 Key Results
* Pre-trained models significantly outperform training from scratch.
* EfficientNet-B0 showed the best performance, achieving 85% accuracy in just one epoch of fine-tuning.
* 🖇️ Try It Yourself
* Google Colab Notebook
Feel free to explore the code and share your suggestions! 🚀