/Inception-Model-Implementation

This repository contains the implementation of the Inception model from scratch and the pretrained V3 model, both used on the flower dataset.

Primary LanguageJupyter Notebook

Flower Classification using Inception Model

Overview

This repository focuses on building and training Inception models for flower classification using TensorFlow and Keras. Two approaches are explored: building the Inception model from scratch and utilizing pre-trained weights from Keras' InceptionV3 model.

Key Feature

1. Dataset

The dataset used for training the models is the Flowers Recognition dataset, which includes images of flowers categorized as [0, 1, 2, 3, 4]. The dataset can be found here.

2. Models

1. Inception Model from Scratch

  1. Architecture

    • Convolutional layers, pooling layers, inception modules, dropout layers, and fully connected layers.

  2. Training

    • Compiled with Adam optimizer and sparse categorical crossentropy loss.
    • 100 epochs, batch size of 256.

2. Inception Model with Pre-trained Weights

  1. Architecture

    • Built using pre-trained weights from Keras' InceptionV3 model.
    • Fine-tuned on the flower dataset.

  2. Training

    • Fine-tuned over 100 epochs, batch size of 256.

3. Learning Curve

  • Plots for training and validation loss for both models.
  • Plots for training and validation accuracy for both models.

4. Evaluation Metrics and Confusion Matrices

  • Generate confusion matrices for testing samples on both custom and pre-trained models.
  • Calculate precision, recall, and F1 score to assess model performance.

5. Results Analysis

  • Explores and comments on the results obtained from custom and pre-trained models.
  • Analyzes the impact of transfer learning on classification accuracy and training efficiency.