Follow Along!

Follow along: Segmentation Examples

The program launching process along with parameter settings are all simplified and set up on the Jupyter Notebook Environment.
The segmentation task is able to operate with various input pictures. Here are the examples:
  • City Scapes
  • Outdoor (off-road)
  • Segmenting Human Images
  • Variaty Objects and People
  • In-Doors
(The Jetson Board used for these examples are => Jetson Nano)

  • 02_5-1. segmentation.ipynb

  • Running the cell code
    Ctrl + Enter

  • Initialize your output stream, and your path, and import in the Image library

from IPython.display import Image

%env DISPLAY=:0
%env PROGRAM_PATH=/home/zeta/jetson-inference/build/aarch64/bin
%env INPUT_PATH=/home/zeta/jetson-inference/build/aarch64/bin/images
%env OUTPUT_PATH=/home/zeta/jetson-inference/build/aarch64/bin/images/test

input_path='/home/zeta/jetson-inference/build/aarch64/bin/images'
output_path='/home/zeta/jetson-inference/build/aarch64/bin/images/test'

CityScapes

  • Check all the available pictures within the system

    !ls $INPUT_PATH/city_*

  • Pick one of the images with urban city scene and initialize the image/ output name.

    image_name = 'ChangeMe'
output_name = 'city_result.jpg'
%env IMAGE_NAME = $image_name
%env OUTPUT_NAME = $output_name

Image(filename=input_path+'/'+image_name)

  • Segment the picture

    %%capture
!python3 $PROGRAM_PATH/segnet.py --network=fcn-resnet18-cityscapes $INPUT_PATH/$IMAGE_NAME $OUTPUT_PATH/$OUTPUT_NAME

  • Show the resulting image

    Image(filename=output_path+'/city_result.jpg')

Outdoor (off-road)

  • Check all the available pictures within the system

    !ls $INPUT_PATH/trail_*

  • Pick one of the images with outdoor scenes and initialize the image/ output name.

    image_name = 'ChangeMe'
output_name = 'trail_result.jpg'
%env IMAGE_NAME = $image_name
%env OUTPUT_NAME = $output_name

Image(filename=input_path+'/'+image_name)

  • Segment the picture

    %%capture
!python3 $PROGRAM_PATH/segnet.py --network=fcn-resnet18-deepscene $INPUT_PATH/$IMAGE_NAME $OUTPUT_PATH/$OUTPUT_NAME

  • Show the resulting image

    Image(filename=output_path+'/trail_result.jpg')

Segmenting Human Images

  • Check all the available human pictures within the system

    !ls $INPUT_PATH/humans_*

  • Pick one of the images with people and initialize the image/ output name.

    image_name = 'ChangeMe'
output_name = 'humans_seg_result.jpg'
%env IMAGE_NAME = $image_name
%env OUTPUT_NAME = $output_name

Image(filename=input_path+'/'+image_name)

  • Segment the picture

    %%capture
!python3 $PROGRAM_PATH/segnet.py --network=fcn-resnet18-mhp $INPUT_PATH/$IMAGE_NAME $OUTPUT_PATH/$OUTPUT_NAME

  • Show the resulting image

    Image(filename=output_path+'/humans_seg_result.jpg')

Variaty Objects and People

  • Check all the available pictures within the system

    !ls $INPUT_PATH/object_*

  • Pick one of the images with variaty of objects and people and initialize the image/ output name.

    image_name = 'ChangeMe'
output_name = 'object_result.jpg'
%env IMAGE_NAME = $image_name
%env OUTPUT_NAME = $output_name

Image(filename=input_path+'/'+image_name)

  • Segment the picture

    %%capture
!python3 $PROGRAM_PATH/segnet.py --network=fcn-resnet18-voc $INPUT_PATH/$IMAGE_NAME $OUTPUT_PATH/$OUTPUT_NAME

  • Show the resulting image

    Image(filename=output_path+'/object_result.jpg')

In-Doors

  • Check all the available pictures within the system

    !ls $INPUT_PATH/room_*

  • Pick one of the images with urban city scene and initialize the image/ output name.

    image_name = 'ChangeMe'
output_name = 'room_result.jpg'
%env IMAGE_NAME = $image_name
%env OUTPUT_NAME = $output_name

Image(filename=input_path+'/'+image_name)

  • Segment the picture

    %%capture
!python3 $PROGRAM_PATH/segnet.py --network=fcn-resnet18-sun $INPUT_PATH/$IMAGE_NAME $OUTPUT_PATH/$OUTPUT_NAME

  • Show the resulting image

    Image(filename=output_path+'/room_result.jpg')