Human Semantic Segmentation

Overview for Semantic Segmentation

For segmentation, demonstrates the steps to use the DeepLab model to perform semantic segmentation on a sample input image. Expected outputs are semantic labels overlayed on the sample image, such as some of examples including.

Case 9 - Overview

Atrous Spatial Pyramid Pooling (1/2)

– It is proposed such as the Atrous Spatial Pyramid Pooling where four parallel atrous convolutions with different atrous rates are applied on top of the feature map.

– The resulting features from all the branches are then concatenated and pass through another 1 × 1 convolution (also with 256 filters and batch normalization) before the final 1 × 1 convolution which generates the final logits.

Atrous Spatial Pyramid Pooling (2/2) 

Case 9 - Atrous Spatial Pyramid Pooling

How to segment every objects at each frames

– Collecting data for training and Calibrating or loading the latest version of the pretrained DeepLab model (supported by Tensorflow)

– Load the colormap from the PASCAL VOC dataset

– Adds colors to various labels, such as “pink” for people, “green” for bicycle, e.t.c

– Visualize an image, and add an overlay of colors on various regions

Experimental Results

Case 9 - Experimental Results