CNN-Tensorflow

Introduction

In this project, I will use tensorflow to construct a Convolutional Neural Network (CNN) for image recognition by using Food-11 dataset.

Dataset

• Food-11 Dataset https://mmspg.epfl.ch/food-image-datasets
• some examples

Implementation

First, I reshape all the pictures to 98x98x3 in order to make training faster. This model is consist of three layers of convolution, three layers of max pooling and two layers of FC.

Results

• learning curve

• some examples of detected and undetected images

• feaure map of detected picture

• feaure map of undetected picture