ARTIFICIAL INTELLIGENCE
HOLIDAYS HOMEWORK
Session: 2025-2026
SUBMITTED TO: SUBMITTED BY:
[Link] REEJUL
9th C
30
SWAMI SANT DASS PUBLIC SCHOOL
Udaseen Ashram ,[Link],Jalandhar
INTRODUCTION
This project explores practical application of Machine Learning in the field of
image classification, specifically focusing on the recognition of various species of
sea animals. The primary objective of this study is to develop an automated
system capable of distinguishing between marine life forms with high accuracy.
To achieve this, a comprehensive dataset of marine imagery was sourced from
Kaggle, a premier platform for data science resources.
Using this data, a custom image classification model was trained utilizing
Google’s Teachable Machine. This project not only demonstrates the
capabilities of modern AI tools but also highlights how technology can be
leveraged to identify and categorize the rich biodiversity of our oceans with high
efficiency. For the specific scope of this assignment, the model focuses on
distinguishing between sharks and whales. By analysing the visual differences
between these two distinct categories, the project aims to showcase how neural
networks can learn to identify unique physical characteristics, serving as a
foundational step toward understanding complex ecological monitoring
systems.
TOOLS USED
For this project, two primary digital platforms were used:
Subsequently, Google’s Teachable
Kaggle and Google’s
Machine was employed to construct the
Teachable Machine. Initially,
actual classification model. This tool
Kaggle served as the resource
facilitated the process of uploading the
for locating and downloading
dataset and "teaching" the computer to
a collection of high-quality
recognize visual patterns without the
photographs featuring various
necessity of writing complex code. By
sea animals. This step
integrating these two powerful tools, a
provided the essential
static set of images was successfully
foundational dataset required
transformed into a fully functional AI
to commence the study.
program capable of image recognition.
OPENING KAGGLE
About Kaggle is a popular website owned by Google where data scientists share
large collections of data. It is an ideal resource for school projects due to its vast
library of high-quality, pre-categorized datasets.
Step-by-Step Procedure
1. Search for the Topic: Navigate to the Kaggle search bar and type in the
desired topic (e.g., "Sea Animals").
2. Select a Dataset: Review the search results and choose a dataset that
features clear, high-quality photos that are already sorted into distinct
categories.
3. Download the Data: Locate and click the "Download" button on the
dataset page. This will save a ZIP file containing the data to the computer.
4. Extract Files: Open the downloaded ZIP file and extract the contents to
retrieve the folders of images, making them ready for the next stage of the
project.
EXTRACTING DATASET
1. Locate the File: Open the
"File Explorer" (on
Windows) or "Finder" (on
Mac) and navigate to
your Downloads folder.
Look for the file you
recently downloaded
from Kaggle; it will have a
zipper icon on it,
indicating it is a
compressed ZIP file.
2. Right-Click the File: Once you have found the correct file, click on it once
to select it, and then right-click your mouse. This will open a dropdown
menu with a list of actions you can perform on the file.
3. Select "Extract All": Look through the menu options and click on "Extract
All". If you are using a Mac, you might just need to double-click the file, or
look for "Open With > Archive Utility." This command tells the computer
to start the process of opening the compressed package.
4. Confirm the Location: A small window will pop up asking where you want
to save the open files. You can usually leave this as the default setting.
Click the "Extract" button at the bottom of the window to begin unzipping
the data.
5. Verify the Data: Open the folder to check the contents. You should see
that the images are already organized into separate sub-folders named
after each sea animal (e.g., "Shark, “Dolphin," "Crab"), which is exactly
what we need for the next step.
SETTING UP THE PROJECT IN TEACHABLE MACHINE
Once the images are ready, the next step is to set up the training model
Step 1: Access the Platform
Navigate to the official website at
the homepage, click the blue
"Get Started" button to launch
the application.
Step 2: Select Project Type
Since this project requires
recognizing visual data, select the
"Image Project" panel from the
available options.
Step 3: Choose Model
Architecture
Select the "Standard Image
Model" option.
Step 4: Configure Categories (Classes)
The interface display’s default boxes labelled "Class 1" and "Class 2".
Rename Classes: Click the edit icon (pencil) to rename these boxes to
match your specific sea animals (e.g., change "Class 1" to "Dolphin" and
"Class 2" to "Shark").
Add More: To include additional animals, click the "Add a Class" button at
the bottom of the list.
UPLOADING IMAGES AND TRAINING THE MODEL
With the classes created, I moved on to feeding the data into the system.
Step 5: Upload Images
For each animal class, click the "Upload" button and select "Choose images from your
files." Navigate to the folder containing the extracted Kaggle dataset and select all the
correct images corresponding to that specific animal. Repeat this process for every defined
class.
Step 6: Train the Model
Once all images are successfully uploaded, click the "Train Model" button located in the
center of the screen to initiate the learning process.
Step 7: Wait for Processing
Allow the system to process the data; this typically takes 1–2 minutes.
Note: Do not close the window or switch browser tabs during this time, as doing so
may interrupt the training process.
Step 8: Verify Completion
Monitor the "Preview" box on the right side of the screen. When it activates (turns on), the
training is complete, and the model is ready to be tested .
EXPORTING THE MODEL
EXPORTING THE MODEL
1. To finish the project and create a shareable link, I exported the
model.
2. Click Export: I clicked the "Export Model" button located above
the Preview box.
3. Select Upload: A pop-up window appeared. I selected the tab that
says "Upload my model" (this saves it to the cloud so anyone can
view it).
4. Get the Link: I clicked the "Upload my model" button and waited
a few seconds for it to process. Once done, a unique link appeared
under "Your shareable link."
5. Save the Link: I copied this link and pasted it into my project
document. This link allows my teacher and classmates to open
and interact with my AI model on their own computers.
TESTING AND EVALUATION
Once the training process is complete, the final step is to rigorously
test the model to evaluate its performance. This phase is critical to
ensure that the Artificial Intelligence can correctly identify sea
animals it has not seen before. The Preview panel in Teachable
Machine is used to conduct these tests by uploading new images
that are not part of the original training set. In most test cases, the
model successfully classifies the species with a high degree of
accuracy (often above 85%), proving that the system effectively
learns the distinct features of each sea animal.
.
OBSERVATIONS
During the testing phase, the Artificial Intelligence exhibited several key
behaviours:
High Accuracy: When clear images of the sea animals were presented, the
model identified them correctly with high confidence levels (often
between 90% and 100%).
Real-Time Response: The model processed visual data instantly. Upon the
upload or display of a new image via the webcam, the probability bars
updated immediately to reflect the new prediction.
Visual Distinctions: The system most effectively identified animals with
distinct morphological features, such as the fin of a shark or the legs of a
crab. Conversely, animals with similar shapes or colours occasionally
resulted in lower confidence scores, demonstrating that the quality and
variety of the training data are critical factors for success.
What I Learnt
The Concept of Supervised Learning: I learned how computers "learn" by example.
By labelling images and feeding them to the system, I understood that the quality
and quantity of data directly affect how smart the model becomes.
Data Management Skills: I learned how to search for specific datasets on Kaggle,
extract files from compressed ZIP folders, and organize them into categories, which is
a fundamental skill in any data science project.
Challenges Faced
Similar Animal Features: One challenge was that some sea animals look very similar
(like different types of fish). The model sometimes got confused between them,
requiring me to find more distinct photos to help it tell the difference.
Data Cleaning: The dataset from Kaggle contained hundreds of images, and not all of
them were perfect. I had to manually filter out blurry or unrelated images to ensure
they didn't confuse the training process.
Relevance in Real Life
Marine Conservation: This type of technology is used by marine biologists to
automatically track endangered species in the ocean using underwater cameras,
saving time compared to manual observation.
Educational Tools: Apps based on this model can be used in aquariums or schools,
allowing students to take a picture of a sea creature and instantly learn its name and
facts about it.