Image_Segmentation_Model

A Fully Convolutional Neural Network based model for Semantic Segmentation

In this project, a segmentation task has been performed for a small sample dataset. The dataset can be downloaded from this link, which has been prepared by Divam Gupta.

A FCN-8 architecture has been used for this task, which uses CNN to convert each of the image pixels to categories.For details of FCN-8, please refer to this paper.

Model Performance

Opensource h5 pre-trained weights of VGG-16 architecture has been initialised at first. The model shows the Training Accuracy of 72% and Training Loss of 0.9