Pentous

An AI that excels in crafting tailored meal and workout plans. Experience Pentous: Your personalized lifestyle app, fueled by your profile.

Comment

Inspiration

Our inspiration for Pentous surrounds the four primary food groups and arguably the most necessary consumable, water. The four primary food groups are ingrained in our memories from elementary school lessons when we first learned not to eat McDonalds daily. The four primary food groups identify healthy eating habits, a system unique to Canadians. The combination of the four food groups and the all-important water forms the spokes that allow the wheels of life to turn. This combination is the origin of the prefix “pent-” while the suffix “-ous” refers to the formation of an adjective, thus describing the five essential elements of nutrition. That concerns the inspiration behind the name of our company; however, more importantly, the reason behind the company's formation lies in the difficulty of living healthily. Many “health” programs exist online solely to exploit those who are aiming to improve their lives financially. We wanted to create a free, easy-to-use service that enables users to take the initiative in their own journey to live healthier lives.

What it does

Pentous allows users to fill out a comprehensive profile about themselves which will build them a personalized lifestyle. To elaborate on lifestyle, this encompasses proper hydration, exercise, sleep, and meal plans. Users can obtain these results after submitting a profile containing information regarding biological gender, height, weight, and existing diet composition. The backend pulls from several APIs that create exercise and recipe/meal plans. A second function utilizing the JART-Classifier enables users to scan their food to determine what nutritional group it belongs to. The program will provide a percentage of the food group that it believes the image belongs to and based on its confidence percentage the user can infer the general composition of their platter.

How we built it

The front end of the Pentous app was built using a combination of HTML5 and CSS3. The combination of these two languages was used to emphasize the healthy aspects of a sound lifestyle. The backend of the Pentous app surrounds a multitude of programs to appropriately generate an application that can complete its given task. To begin, machine learning coded with Python was implemented to create the JART-Classifier (Journey to Attainable Results Together). This classifier allowed users to scan images of their food to see what food groups their meals were composed of. Furthermore, a manual input system was implemented through the use of Next.js and Typescript. Using the Reactjs framework was much better.

Challenges we ran into

With machine learning, we originally attempted to implement it to find the specific percentages of each nutritional group in a designated image. However, the program was very often wrong and was frequently inaccurate. This was due to the very little amount of layers we were able to add to the Sequential() model. Using only 3 layers of relu on Conv2D was not very great for making an extremely accurate model. The amount of time to train was much shorter though.

Connecting to our API's were also quite troublesome, as we switched from jQuery to axios. the axios library was easier to import and use the get request to retrieve our data. This resulted in lots of lost time for documentation and understanding jQuery when we ended up switching.

The CSS folders were not quite organized, alongside the tsx folders. Overtime, we got sloppy and lost our organization with good practices. Even more, using tsx and a framework was not great as we just learned vanilla js. However, we figured it was much easier to start from a framework.

Accomplishments that we're proud of

One of the several accomplishments that we are proud of is our relative inexperience concerning many of the tools that were used throughout this project. For instance, HTML and CSS were primarily learned after signing up for the competition along with the basics of machine learning. Moreover, the fact that our team was able to properly set up a website with a legitimate web domain that genuinely works is absolutely astonishing to us. Seeing and witnessing the actual progress of our work throughout the 48 hours was a special experience for our team, and we were proud to see our final product. Additionally, the entire experience of the RecessHacks hackathon was a collaborative experience, and we were very proud to work as a team despite all the hurdles we were forced to overcome. Due to the limited period of time, it was essential for us to appropriately designate and divide our timeline. Originally we created a plan that we thought was sound but as a result of the hurdles that we suffered above it had to be altered many times. Despite this, our collaborative efforts and overall flexibility throughout our team enabled us to maximize our productivity and to still produce a product that we are very proud of at the end of the day. Even more, we are proud that our machine learning model ended up training to a good spot and was successful in classifying typical inputs. It was a novel thing for us to follow and learn, so just getting it to train was very exciting!

What we learned

Throughout this entire process, we learned an expansive array of new skills. For instance, due to many of the hindrances we suffered above, I learned to be far more flexible with changing the original plan and design of the project. Moreover, through the preparation and actual execution of the project, some of our group members learned the basics of HTML and CSS. This opened the door to the world of front-end development, which added to our prior back-end development skills. Another skill that we learned was determination and collaboration. Late nights and intense coding sessions became the norm as the deadline approached. Despite the challenges, our team's camaraderie and shared commitment retained our motivation to push through. During the hackathon, we gained valuable insights that transcended technical skills. Collaborative teamwork underscored the significance of effective communication and task delegation. Time constraints honed our ability to manage tasks efficiently. Presenting our project refined our ability to convey complex ideas. Most importantly, we realized the purpose and impact of our work, recognizing technology's potential for positive change in the real world.

What's next for Pentous

For Pentous our future goal is to improve our program. Currently, there are a multitude of issues with our program. The original intention of our machine learning program, JART-Classifiers was to identify the percentage of each nutritional group present in a particular food platter. However, the program constantly failed to work and could not accurately identify the portion of each nutritional group in a given image. Moreover, for a future version of Pentous, we would like to add more functionality to it. This would include a system that would identify rotten food. Additionally, we would like to improve the personalization and accuracy of said system. Currently, the personalization of the Pentous system is quite broad and does not cover many important aspects of life that could influence one’s success in a particular program. These range from income level, work hours, and current lifestyle. Furthermore, rather than simply providing a workout based on the level type of the user, our team can put in place a more accurate system with more questions to optimize how effective the user’s workout program will be. Questions that could optimize the workout range from: how many times do you workout a week, how long are your workouts, what accessibility do you have to a gym, where do you prefer to workout etc.

Built With

Share this project:

Updates