Fruit Detection Model - Machine Learning Project

Project Overview

I developed a custom fruit detection model using a dataset of 100 fruit images that I captured and manually labeled with Label Studio in an Anaconda virtual environment. I trained the model in Python on Google Colab with a T4 GPU for 60 epochs, then deployed it locally on my RTX 4060 GPU with a 720p webcam for real-time testing. To run the model on my machine, I configured an Anaconda environment, installed the proper drivers and the latest version of PyTorch, and successfully achieved live fruit detection through the command prompt.

Key Features

• Runs Locally
• Pytorch
• Anaconda
• Label Studio

What I Learned

Through this project I learned how to create my own data set, I payed close attention on how the images came out I tried using different backgrounds and different lighting in order to have a diverse pool of data to train the model. I also learned how to label my data efficiently through Label Studio which was a long task, after labeling my data I worked with the YOLO framework, I chose YOLO 11 because I had capeable hardware to run it. to write all of my code I used a Google Collab notebook, beucase it allowed me to train the model using a powerful T4 GPU through the cloud, it was also useful because it provides a virtual environment to work with, I also gathered some data on how the model was performing through the training, the graphs show the models loss and other important data. After the model was fully trained I deployed it on my own machine using an Anaconda virtual enviroment with python 3 and the lastest version of PyTorch.