Semantic segmentation with UNET

Original Image	Predicted Segmentation Image	Image with segmentation mask

This project focuses on implementing semantic segmentation on pascal VOC dataset using UNET. Implementation of UNET is achieved with fastai library. Accuracy achieved is 78.75% with output image size 224x224. GPU used for this project: NVIDIA GeForce GTX 1050

DATASET

---

Pascal VOC 2012 challenge dataset is popular for object detection and segmentation task. This dataset consists of images which belong to 20 different types of objects. Therefore, 21 classes for predictions including background(class index=0):

Person: person
Animal: bird, cat, cow, dog, horse, sheep
Vehicle: aeroplane, bicycle, boat, bus, car, motorbike, train
Indoor: bottle, chair, dining table, potted plant, sofa, tv/monitor

Class indices for semantic segmentation task:

0=background, 1=aeroplane, 2=bicycle, 3=bird, 4=boat, 5=bottle, 6=bus, 7=car , 8=cat, 9=chair, 10=cow, 11=diningtable, 12=dog, 13=horse, 14=motorbike, 15=person, 16=potted plant, 17=sheep, 18=sofa, 19=train, 20=tv/monitor

more example on semnatic segmentation on pascal voc dataset = http://host.robots.ox.ac.uk/pascal/VOC/voc2012/segexamples/index.html

Downloading dataset:

Download the development kit for training/validation data. More information

http://host.robots.ox.ac.uk/pascal/VOC/voc2012/#data

Alternative resource of downloading the dataset is using:

https://s3.amazonaws.com/fast-ai-imagelocal/pascal-voc.tgz

Directory structure:

After downloading the development kit following will be it's folder structure:

VOCdevkit/VOC2012
VOCdevkit/VOC2012/Annotations
VOCdevkit/VOC2012/ImageSets
VOCdevkit/VOC2012/JPEGImages
VOCdevkit/VOC2012/SegmentationClass
VOCdevkit/VOC2012/SegmentationObject

Libraries

fastai
numpy
jupyter notebook

Outputs:

input and out images

Input Images	Predicted Segmentation mask	Image with segmentation mask