SIMBA – Software with Intelligent Marking-Based Identification – Case Study

  • Home
  • Case Studies
  • SIMBA – Software with Intelligent Marking-Based Identification – Case Study

How we made an AI based software for recognition and accurate counting of the lions of Gir Forest

Project Description

Software with Intelligent Marking-based identification of lions (SIMBA) is an AI-based software to identify lion species accurately through facial recognition. The software is developed by training deep learning models to fetch facial features and recognize each individual lion accurately. A key trait of the software is that it maintains a database of lions and their medical history. Further, the client can track their lions through geolocation and keep tabs on dead and alive lions.

Business Objective

Population estimation of wide-ranging carnivores has always been a challenge to wildlife managers. In the case of endangered carnivores, population estimates need to be precise and accurate since a small decline in such a population could prove disastrous.

The Teliolabs team set out to develop an AI-based automated lion recognition and counting system.

Pain Points

Since 1963, the Gujarat forest department has performed a census every 5 years. For the 1995 census, lions were baited with live domestic buffaloes for three consecutive days throughout the entire lion range (over 1). Over 250 buffaloes were used and about 1,800km 21500 man-days consisting of forest staff and volunteers were employed in conducting this massive census operation.

With the project, the client wanted to eliminate the following pain points:

  • The tedious process (like pugmarks, track counts, scent plots, mark-recapture, radio telemetry, and other estimation methods) of counting lions manually
  • Inaccurate count due to human error factors like duplicacy
  • Costly due to various resources involved
  • Hunting and poaching of lions
  • Lions missing from national parks
  • Illegal movement of lions

High-level solution framework used to resolve the issue

The user interface was developed with angular js, using an ionic framework. Moreover, the backend APIs were created with python. The solution functions by performing image processing on the lion images uploaded to the system. Different AI algorithms, such as binary classifier to identify lion, facenet to match lion image with an existing database, and whisker spot detection are used.

The images are saved in the database and can be viewed in the lion details section. In addition, the application is integrated with Google Maps to enable it to locate lions on a map. Another advantage is that the medical history of lions can be downloaded in excel format.

Tech Stack

After garnering the requirement, the Teliolabs team employed the following tech stack:

  • Frontend – Angular with ionic framework
  • Backend – Python
  • Database – PostgreSQL
  • Deep learning models – binary classifier, facenet, CNN, whisker pattern matching

SIMBA is created by implementing the best standards of different technologies such as:

  • Python
  • Angular with Ionic Framework
  • DataBase
  • Postgres
  • Flask for API development
  • Deep Learning Models
  • Tensorflow
  • Pytorch
  • Google Cloud Platform for:
    • Services
    • Deployment

Key business benefits delivered by Teliolabs

The SIMBA project yielded:

  • An accurate solution for automatic counting and tracking of lions
  • A solution that supports both cloud and on-premise deployment

Project Niche

The project, being one-of-a-kind in India, was a major milestone for the company in the artificial intelligence/machine learning niche.

End Customer

Gir National Park, Gujarat, was the end customer for the SIMBA project.

Achievements

Notably, the team employed the binary classifier model to bolster performance.  The accuracy of the AI model consequently improved 3-fold, from 30 to 90%.  Moreover, whisker spot detection was also achieved.