You are on page 1of 8

1.

Group Id : 6

2. Project Title
Wildlife Conservation and Analysis using Machine Learning

3. Project Option
Internal Project

4. Internal Guide
Dr. Mahendra Salunkhe

5. Sponsorship and External Guide


None

6. Technical Keywords (as per ACM Keywords)


● Kalman Filter
● Hamilton and Euclidean formula
● SVM
● RandomForest
● DecisionTree
● CNN
● YOLO

7. Problem Statement
A web application that counts the number of animals in a specific region
using camera recordings, classifies animals as endangered or extinct using
images, and summarizes animal information from across the world,
including migration and biodiversity of animals in various national parks.
8. Abstract
In the realm of wildlife conservation, the preservation of Earth's rich
biodiversity is an increasingly urgent imperative. Our project is dedicated to
advancing this vital cause through a multifaceted approach. We employ
cutting-edge technologies, data analysis, and classification criteria to address
critical aspects of wildlife conservation.

This wildlife conservation project presents a holistic approach comprising


three core components:

● Calculating Number of Animals per Frame:

Leveraging cutting-edge computer vision and machine learning techniques,


we quantify wildlife populations accurately from images and videos. This
data-driven approach streamlines data collection and monitoring efforts,
enabling precise population assessments crucial for conservation.

● Summarizing Information about Animals from Across the World:

We compile extensive information on animal species from diverse global


sources, creating comprehensive profiles. These profiles encompass crucial
data, including species distribution, population trends, habitat characteristics,
and identified threats. This unified dataset offers a profound understanding
of the world's wildlife.

● Classification of Animals as per Red List Criteria:

Applying the rigorous criteria established by the IUCN Red List, we


categorize species based on their endangerment levels. By considering
population size, habitat loss, and threat assessments, we provide vital
insights for informed conservation decision-making.

This integrated framework facilitates data-driven wildlife conservation,


empowering stakeholders with accurate population data, comprehensive
species insights, and a standardized approach to assessing conservation
priorities. By combining technology, data analysis, and conservation science,
we aim to contribute significantly to the protection and preservation of the
Earth's diverse and invaluable biodiversity.

9. Goals and Objectives


a. For accurate wildlife population counts from images and
videos, it is important to develop tools that incorporate
computer vision and machine learning.
b. Create species profiles that compile detailed information on
population trends, habitat characteristics, identified threats,
and distribution data.
c. For informed conservation decisions, it is important to
categorize species according to the IUCN Red List criteria
for endangerment classification.
d. Tailored conservation strategies are aided by data collected
on animal migration and biodiversity in national parks,
providing insights into biodiversity.
e. Develop a regularly maintained unified platform to
effortlessly access and combine data.
10. Relevant mathematics associated with the Project
System Description
● Input
Sequence of images
● Output
There are three parts to our solution
○ Calculating number of animals per frame
○ Summarizing information about animals from
across the world
○ Classification of animals as per Red List criteria
● Data structures
○ Lists
○ Numpy Arrays
○ Pandas Dataframe
● Functions
○ Lambda Function
■ To manipulate and process data
● Possible Algorithms to be Used
○ Calculating number of animals per frame
■ Hamiltonian Formula
■ Euclidean Formula
■ Kalman Filter
○ Classification of animals as per Red List criteria
■ This involves two parts:
● Detection
○ CNN
○ YOLOV8
● Classification
○ Random Forest
○ Decision Tree
○ SVM
○ Naive Bayes
● Success Conditions
○ Calculating number of animals per frame
■ The number of animals are correctly counted
○ Classification of animals as per Red List criteria
■ The input image of animal is correctly
identified and classified
● Failure Conditions
○ Calculating number of animals per frame
■ Animals incorrectly counted
○ Classification of animals as per Red List criteria
■ No input image is given
■ Animal wrongly identified
■ Animal incorrectly
11. Review of Conference/Journal Papers Supporting Project idea.

Sr Author Title Advantages Limitations


No.
M. A. Thalor, R. Deep learning Many different . As classifying the mental
Nagabhyrava, K. insights and techniques and health data is generally a
Rajkumar, A. methods for algorithms had been very challenging problem,
Chakraborty, R. classifying wildlife introduced and the features used in the
1. Singh and U. proposed to test and machine learning
Singh solve the mental health algorithms will
Aswal,[2023] problems. significantly affect the
performance of the
classification
M.Mangale- A Comparison of The paper presents a Effective wildlife detection
swaran and M. Different Learning reliable method for method in conservation
Azhagiri [2023] Algorithms for wildlife detection and remains unclear.
Wildlife Detection classification using
2. and Classification various learning
in Animal algorithms, addressing
Conservation critical conservation
Applications. needs.
N. K. Jana, A. Classification of The Model's high accuracy may
Chowdhury, D. Animal Images "MODIFIED CNN" not generalize well to all
Saha, C. Dutta, with "MODIFIED deep learning model animal species
and A. Maiti CNN" model. demonstrates high and
Oct. [2022]. accuracy (95%) in May require careful
classifying animal parameter tuning and could
3. species, which can aid potentially introduce noise
in efficient wildlife or artifacts into the
monitoring and classification process.
preservation efforts.

Tuia, D., Perspectives in ML integration Analyzing large, diverse


Kellenberger, B., machine learning improves ecological sensor data efficiently
4. Beery, S. et for wildlife understanding and demands significant
al.[2022] conservation. modeling through computational resources.
enhanced data
processing.

Duhart, Clément Deep Learning for The Tidzam Deep Deploying deep learning
& Dublon, Wildlife Learning framework systems in remote and
Gershon & Conservation and offers the potential to hostile environments, such
Mayton, Brian & Restoration Efforts. automate wildlife as rainforests or
5. Davenport, detection, identification, high-altitude regions, can
Glorianna & and geolocalization, be technically challenging
Paradiso, Joseph. aiding in large-scale and may require
[2019] monitoring and self-sufficient energy and
conservation efforts. computational solutions.
Aarju, R. Enabling Its provided that challenge of emotion
Bahuguna, S. Technologies for comprehensive survey fusion is still there
Pandey, R. Wildlife of how emotion fusion
6. Singh, H. Kaur Conservation. has been used to
and G. Chhabra support this task
[2023]
Miao, Z., Insights and Intelligent software for Deep learning models may
Gaynor, K.M., approaches using wildlife classification lack interpretability,
7. Wang, J. et deep learning to offers efficient and making it challenging to
al.[2019]. classify wildlife. accurate species understand the rationale
identification, behind their classifications
enhancing conservation fully.
efforts.

You might also like