Using Deep Learning AI Models to Help a Private Owned Company Safeguarding Environment & Species

A privately funded body was assigned with the responsibility of setting up a specific watershed stewardship plan that helped a specific regional tribe in the US to help manage and monitor different species of fishes. Initially, they were using an old legacy system via hardware support in the form of infrared sensors and video cameras that meant breaking the bank.

 

To overcome the unwanted monetary costs, they approached Glorious Insight to provide a solution using technology. We used predictive analysis, AI deep learning models, and AI-driven image recognition systems that helped reduce more than 5 times the cost incurred by installing the legacy systems,i.e., the hardware supports like sensors, HD cameras. 

 

Customer Background 

 

The client has been working towards analyzing, assessing, and managing the nature via specific nature conservation programs. In one of their projects, they had to assist locals in the US to help manage different species of fishes that swim across the river for analyzing their behavioral patterns and counts. Though everything was going perfect for the privately funded body, they wanted a different technology approach that can help them stay innovative from a technology point of view and save cost at the same time. 

So, they collaborated with Glorious Insight to help them identify patterns of the fishes swimming across the river and predict the species that are less in number and they need a revival plan for restoration in that new habitat. 

 

Business Challenge 

The key challenge for the privately-owned river foundation was to shift their dynamics from fully-fledged hardware equipped infrastructure to software-driven economical scalable solutions. In this pursuit, they wanted someone who could help them explore the potential of data to understand patterns and predict the species that have specific time slots that they prefer the most to swim across in the river. 

 

Solution 

 

Glorious Insight took cognizance of the situation and went through all the videos that were recorded so far to understand the patterns of the fishes swimming in the river for classification. They adopted Deep Learning AI Models that drew lines around each species of fishes that swam in the river. All the developments in the way they swam in the river at different times were converted in a work-flow embedded in an application that simply automated the process where video feed inputs, detection, and classifications could all work simultaneously. To make the process work, automated AI solutions were put to use that worked upon deep learning models using Azure & Cognitive Services Platform Stack. 

 

Contribution of Glorious Insight To Solve the Problem 

Processing all the videos for an actionable insight demanded significant manual work.

  • To help overcome the manual intervention that could have cost fortunes in the form of salaries and equipment, we automated the process via VOTT tools. 
  • All the frames that we tagged for fish identification were used to train a basic software model using CNTK  (Cognitive Tool Kit & Faster RCNN ( Regional -Convolution Neural Network). But while resolving the problem, we faced concerns regarding speed and real-time videos when more frames and videos were put to use. 
  • To mitigate the video analysis problem, advanced object detection systems with faster identification protocols were integrated. 

 

Quantifiable Outcome of Deployment 

 

The web-based AI solutions that we have deployed for the privately funded body had helped them save significant man-working hours of biologists and cost input for managing the infrastructure. The outcome was a significant 80% drop in the expenditure incurred for classifying, analyzing, and restoring data from fish assessment exercise.