True Random Number Generator with comprehensive statistical tests

bluedot-tile

DeepField-SR

DeepField-SR is a fixed functional hardware accelerator leveraging Xilinx Alveo Cards to offer the highest computational efficiency for Video Super Resolution. Based on proprietary Neural Network trained with real world video data from the internet and fusing spatio-temporal information in multiple frames, it produces superior high resolution video quality.

Vendor: BLUEDOT

Last Update: March 15, 2021

Size: 706MB

Container version: DeepField_SR_v1.0.0

Try or Buy

Obtain an entitlement to evaluate or purchase this product.


Begin a free trial and run the application example below.


Deployment Options

This application is containerized and can be easily run in a few minutes in the cloud.

Cloud Services
AWS
  • Xilinx Runtime: 2020.1
  • AWS AMI:  DeepField-SR
  • Target Platform: f1.2xlarge

 


Start Evaluation

Follow the instructions based on your deployment method.

AWS

1.

Obtain Evaluation Access Key

Create and download the cred.json file from the Xilinx App store (https://appstore.xilinx.com/).

  • Login to the Xilinx App Store Portal
  • Click the button labeled "Manage Account" to view entitlements.
  • Click the "Access Key" link on the left side menu
  • Click the "Create an Access Key" button.
  • Download the resulting file "cred.json"

2.

Setup AWS

Access the AWS EC2 console and allocate an F1 instance using the procedure below.

2.1 Select DeepField-SR AMI

Execute DeepField-SR AMI

  • Go to the AWS console (https://aws.amazon.com/en/console/) page and log in
  • Change the region to US East (N. Virginia) us-east-1 and click on EC2 in the Compute category
  • Select AMIs from the Images category
  • Select the search box category as Public images and search for DeepField-SR
  • Right-click on the search result and click "Launch"

2.2 Select an instance type

Select f1 for the instance family and f1.2xlarge for the instance type as shown below. Then click the "Review and Launch" button at the bottom.


2.3 Create a private key

When the "Launch" button on the Review page is clicked, a pop-up to register or create a private key appears. If private key already exists, select it from the dropdown menu. Otherwise, create a new one.

Below is an example of the process of creating a new one.

  • Choose "Create a new key pair" from the drop-down menu.
  • Enter an appropriate name in "Key pair name".
  • Click the "Download Key Pair" button and save your private key in a folder.
  • Click the "Launch instances" button.

2.4 Run DeepField-SR

  • If the f1 instance created in the EC2 console is selected, IP address can be checked in the detail pane right.
  • Change the private key created in ‘1 Obtain an Evaluation Access Key’ to read-only permission.
  • In the terminal program, connect with the private key as shown in the command below.  
    $ chmod 600 <your_download_folder>/deepfield-sr.pem
$ scp -i <your_download_folder>/deepfield-sr.pem <cred.json> ec2-user@<instance_ip_address>:/home/ec2-user/
$ ssh -i <your_download_folder>/deepfield-sr.pem ec2-user@<instance_ip_address>



######                       #######
#     # ###### ###### #####  #       # ###### #      #####
#     # #      #      #    # #       # #      #      #    #
#     # #####  #####  #    # #####   # #####  #      #    #
#     # #      #      #####  #       # #      #      #    #
#     # #      #      #      #       # #      #      #    #
######  ###### ###### #      #       # ###### ###### #####

                                        https://blue-dot.io
                                           info@blue-dot.io
                                              Trial version

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Try our SR solution

# Loading the AFI image
$ load_afi

# test using the ffmpeg
$ sudo ./ffmpeg -i test_720p.mp4 -c:v libx264 -b:v 20M -vf scale_startrek=w=iw*3:h=ih*3:fpga=f1 output_4k.mp4

## for more informations of scale_startrek
$ ./ffmpeg -help filter=scale_startrek
The option "c" isn't supported in a F1 instance. Please don't use "c" option.


# test using our stand-alone program
$ sudo ./deepfield -i test_720p.yuv -w 1280 -h 720 --dw 3840 --dh 2160 -p f1 output_4k.yuv

## for more detail information of options
$ ./deepfield -h

# test clips
- test_720p.mp4 : 1280x720, VP9, 1 minute
- test_720p.yuv : YUV420 8bit 500 frames
--------------------------------

Note
Max input size : 1280x720
Max output size: 3840x2160

--------------------------------




$ sudo ./ffmpeg -pix_fmt yuv420p -s <input_width>x<input_height> -i <input_path> -vf scale_startrek=w=iw*3:h=ih*3:fpga=f1 <output_path>


When connected, the banner is displayed. The banner describes how to upload an FPGA image and how to use it. There are two programs provided, ffmpeg and deepfield.


FFMPEG

When using ffmpeg, use scale_startrek as the video filter. The options for scale_startrek filter are as follows:

  • w : target width
  • h : target height
  • fpga : target FPGA platform
  • c : number of FPGA

Note : Option ‘c’ is not available on f1 instance

The option "scale_startrek=w=iw3:h=ih3:fpga=f1" used for banner is the size of the input image (iw, ih) set to 3 times the output size.


3.

Results

deepfield

The deepfield program receives YUV420 8bit video, scale-up, and saves the result in YUV420 8bit.

options

  • w : width of the input image
  • h : height of the input image
  • dw : width of the output image
  • dh : height of the output image
  • p : FPGA platform.
  • n : number of frames

The same can be run with ffmpeg. The example below is the ffmpeg command that scales up YUV and saves the result in YUV.

    $ sudo ./ffmpeg -pix_fmt yuv420p -s <input_width>x<input_height> -i <input_path> -vf scale_startrek=w=iw*3:h=ih*3:fpga=f1 <output_path>

The deepfield program is helpful when checking only the performance of DeepField-SR.