Garrett
James
Josh
Guy
Nicholas
Cristian
About The Project
A neural network application for categorizing different fruits within an image and identifying their ripeness.
Users will be able to submit any image to the application, which will then identify fruits the image may contain. The type, ripeness, and approximate quantity of visible fruit will be given to the user.
The application will be accessible to users through a web UI, to which users may upload an image and receive a list of fruits the image contains, along with information about it.
| OS | Linux, Windows, or Mac |
| GPU | CUDA-Capable GPU (highly recommended, but optional) |
| Package Manager | Pip, Conda, or LibTorch |
| Programming Language | Python 3.x or C++/Java |
| Machine Learning library | Pytorch |
| Graphing, Analytics, & Statistics library | Plotly and/or Pandas |
| Misc. | Dataset (pictures) and Templates (Neural Network models) |
Demand Analysis
Identify Market:
The market for the product would be the set of users for which they do not know the name, number, or ripeness of a set of fruit. The most likely market for this project is as a tool for customers in a supermarket. This project could also be useful for people who have a home grown garden and have encountered a fruit in their garden that was not supposed to be there.
Identify Usefulness:
In a Supermarket environment the project could be used to identify the ripeness of a particular set of fruit. It could also be used to identify the fruit should there not be a sign to tell the customer what the fruit is. If an unknown fruit has popped up in a garden this could help the user determine if the plant is safe. This could be useful because the ripeness of a fruit has an impact on its taste and edibility.
Identify Competitors:
There are a great many different variations on using machine learning to be able to identify fruits and vegetables. There are also very many different machine learning programs for being able to detect the ripeness in fruit. Also being able to detect how many objects are via machine learning is a very common trait. However as this is a project that is just meant for school the competitors will not end up meaning much, but if it were to actually go into production the competition would make success very difficult.
Feasibility Analysis
Technical Feasibility:
To create this project, the team will need software, package-managers, programming libraries, and a sizeable dataset of pictures containing a variety of fruits. A CUDA-capable GPU is optional but highly recommended as it can drastically improve performance. All of the required materials can be sourced on the Internet at no cost to the team or prospective clients. GPUs with CUDA capability cost anywhere from $200 (for used or older models) to over $10,000 (for newer, high-end models).
Financial Feasibility:
All the resources for this project, aside from time, are readily available with no cost to team or prospective clients. In a real-life scenario, this would be different. The project would need to include the pay-rate for the team, cost of necessary hardware, cost of necessary software (if commercial license is necessary), and the cost of installation. Additionally, rates would need to be established for any extra time and materials initially unforeseen.
Organizational Feasibility:
Team consists of 6 members with experienced Computer Science backgrounds. Each member has different areas pertinent to the project where they excel. Collectively, the team has experience with GitHub, Python, Java, C/C++, Web Development, Project Management, and Software Engineering. Every aspect of the project falls within the range of the group.