Edge AI and Video Analytics Solution Provider

Phelps2020 is a women-owned small business founded in 2009 and headquartered in Knoxville, TN. It specializes in artificial intelligence, data fusion, image and video processing, exploitation, and dissemination. Phelps2020’s engineers have decades of experience integrating sensors, specialized hardware, optics, processing, artificial intelligence, visualization, and dissemination of video data in real-time and at the edge. Phelps2020’s innovative solutions address all types of Electro-optical, infrared, and multi-spectral sensors and are applicable to a spectrum of government and commercial use cases. 

Design and Integration Services

Phelps2020 began its pursuit for superior machine learning and computer vision solutions in academia, found a need for its capabilities in the federal and military space and is now providing answers to a multitude of commercial entities in need of artificial intelligence and computer vision solutions.

Our scientists possess a unique depth of expertise across a range of specialties rarely mastered in combination: Optics, Hardware Architecture, Artificial Intelligence, Computer Vision and System Integration.

We are passionate about using this expertise to transform video into actionable information for our clients.

Technology
Advancements

For both internal development and acquisition, Phelps2020 provides its technical knowledge in the form of feasibility assessments, technology roadmaps, market surveys and industry reviews, ensuring that clients apply their resources effectively.

Algorithm and
Hardware Design

Phelps2020 helps clients solve unique video analytics issues through services such as custom developed algorithms, intelligent integration os existing software and hardware (FPGA, GPU, CPU), and configuration of industry solutions.

Databasing, test,
and evaluation

Phelps2020 provides data collection services and protocols’ set-up in various environments. For existing video and AI solutions, we also provide testing and evaluation services which benchmark solutions against industry standards to identify deviation and confirm claims.

Existing Solutions

Hardware

Edge-AI Platform

  • Edge-AI Platform is a low size, weight, and power versatile hybrid hardware solution that ingests 6 different types of video inputs in the visible and/or Infrared, mitigates multiple concurrent degradations and corrects for jitter and atmospheric distortions for long range horizontal paths
  • Edge-AI Platform then applies optimized AI for the detection of people, vehicles, maritime vessels, and airplanes also in the visible and Infrared bands. Edge-AI also matches and recognizes faces, license plates, hull numbers, and tail numbers.
  • Edge-AI Platform provides an FTP service and streams out AI-driven smartly encoded video in H264 and H265, accommodating bitrates as low as 64kbps while preserving high quality regions of interest.
  • Edge-AI Platform is modular, scalable, and customizable to smaller targeted systems like Crystal Edge and 5K-TVS Edge.

5K-TVS Edge

  • 5K-TVS Edge is a small form, hardware solution (FPGA) that accepts video signal as input and returns a corrected version for long range (up to 5 km) visual threat detection and subsequent processing
  • 5K-TVS Edge merges two dozen different functions to intelligently correct a spectrum of video quality issues, concurrently and in real-time, adding to Crystal atmospheric stabilization for long range video
  • 5K-TVS Edge provides an accompanying tablet and GUI for minimal interaction
  •  

Crystal Edge

  • Crystal is a plug-n-play, small form, hardware solution (FPGA) that accepts video signal as input and returns an enhanced version for viewing or processing
  • Crystal merges a dozen different functions to intelligently correct a spectrum of video quality issues concurrently and in real-time
  • Crystal requires no configuration as it moves across different environments
  •  

Software

AI Platform App

GPU-accelerated Windows and Linux applications that perform the following functions in the visible and Infrared bands:

  • Accepts and sends multi-protocol uni- and multi-cast streams
  • Accepts USB cameras
  • Accepts still images and pre-recorded video sequences
  • Mitigates multiple concurrent video degradations
  • Corrects of Atmospheric Distortions and Jitter
  • Detects of People, vehicles, boats, and airplanes
  • Recognizes faces, license plates, hull numbers, and tail numbers
  • Adaptively encodes outgoing streams to accommodate low bandwidth while preserving quality in important regions
  • Records video upon detection
  • Saves and FTPs stills upon detection

5K-TVS App

GPU-accelerated Windows and Linux applications that perform the following functions in the visible and Infrared bands:

  • Accepts and sends multi-protocol uni- and multi-cast streams
  • Accepts live USB cameras
  • Accepts still images and pre-recorded video sequences
  • Mitigates multiple concurrent video degradations
  • Applies Super-resolution
  • Corrects of Atmospheric Distortions and Jitter

Crystal App

  • Crystal App is a standalone GPU-accelerated desktop application (Linux, Windows, and Android) that accepts live and pre-recorded videos and still images
  • Crystal App performs automatic hands-off enhancement, mitigating low light, extreme contrast, washout, backlighting, fog, haze, dust, and pollution
  • Crystal App performs super-resolution and denoising
  • Crystal App is also an API plug in that integrates with existing video viewing systems to automatically enhance images, video clips, and video streams in real-time

Results: Stills

VISIBLE ENHANCEMENT

filefile (1)
file-4file-3
file-2file
file (2)file (3)
filefile-2
file-5file-6
file-7file-8
file (4)file (5)
file (6)file (7)
filefile (1)

INFRARED ENHANCEMENT

VISIBLE-AI

INFRARED-AI

Results: Videos

SUPER RESOLUTION + ENHANCEMENT

ATMOSPHERIC CORRECTION

JITTER MITIGATION

VEHICLES AND PEDESTRIANS DETECTION

ADAPTIVE COMPRESSION

Traditional H264 compression of 1280x720p@15fps and 128 kbpsAI-powered H264 compression of 1280x720p@15fps and 128 kbps
Traditional H264 compression of
1280x720p@30fps and 128 kbps
AI-powered H264 compression of
1280x720p@30fps and 128 kbps

Clients & Partners

error: Content protected