Image Classification, Keras & Tensor flow

Jason Ismail

Apr 5, 20222 min read

I have a strong desire to pursue a career that utilizes Neural Networks. For this project, I gathered my own data using a remote Raspberry Pi that collected image data for microgreen plants. The project was built for a business in Arizona called Thrive and grow farms. The goal was to build a neural network that could differentiate between microgreen plants. For this project, we selected two popular microgreen plants speckled peas and black oil sunflowers.

My Final Presentation can be found here:

Here is an interview with the owner of Thrive and grow Farms after the work was completed:

Here was my project proposal for the class. You can see I like to be organized when planning a project. This project was very unique and designed from the ground up by me while collaborating with the owner of Thrive and grow farms.

This project featured a neural network that was highly accurate at distinguishing between the different microgreen plants.

“The Project was able to accurately predict microgreen images with a 98% success rate”

Key Features for this project:

All data was gathered using a raspberry pi 3B+.
Data was collected using USB cameras.
Relays were employed to control the lighting system remotely.
I built a new computer by hand for this project using a AMD 5900x CPU and a powerful GPU.
The network was built locally using a high-powered Nvidia 3090 RTX GPU.
The project also ran using WSL2 without GPU support as well as docker. (Abandoned for the GPU support in Linux)
The project was built in Linux Ubuntu 20.04 for GPU support.

To optimize the network I built a Jupyter Notebook that could build Neural Networks with different settings and save the networks to file.

It featured an early stopping system that served two purposes. One early stopping could save on processing time allowing me to make predictions faster with my neural network. Two it allowed me to avoid over-fitting the network to the data.

I also built a Network loader notebook that could load any of the previous networks I had built and make new predictions. This helped me optimize my neural network.

This project was designed and implemented by Jason Ismail. A student at Bellevue University. All intellectual rights reserved.

Feel free to contact me if you have any

Comments

Commenting has been turned off.

Looking to Hire?

Connect with a Versatile Data Scientist

Are you in need of tailored data science solutions for your business? I'm here to help. With a Master's Degree in Data Science and a Bachelor's in Mathematics, I bring a blend of academic rigor and practical experience to the table.

Expertise in Building Comprehensive Data Solutions:

Proficient in developing end-to-end data science projects, including the collection, cleaning, and analysis of raw data.
Specialized in Python.

Technical Proficiencies:

Skilled in using Pandas, Yolo, NumPy, PyTorch and Keras/TensorFlow for creating sophisticated Deep Neural Networks.
Experienced in computer vision and leveraging Nvidia CUDA for high-performance computing tasks.

Personal Qualities:

Recognized by peers, mentors, and students as a dedicated and hardworking professional. I come with a long list of references.

Known for facing challenges head-on and being a supportive team player.
Skilled at making complex concepts accessible and relatable, with a passion for continuous learning.

Contact Information:

Jason Ismail
Masters in Data Science, Bachelors in Mathematics
LinkedIn Profile
Phone (Text Only): 719-322-8479

About Me

Data Science

Data Science isn't just my career; it's the realization of a lifelong passion where my love for mathematics, programming, and technology converge. Over the past 20 years, I've nurtured a deep fondness for computers, starting from building them to exploring their immense capabilities.

My academic path initially led me to programming and then chemistry, where I excelled nationally in the 98th percentile. This experience, however, led to an epiphany - it was the mathematical elements within chemistry that truly captivated me. This revelation steered me towards a scholarship in Mathematics and a subsequent career in teaching.

But the true calling came with Data Science. Here, I found an exhilarating opportunity to transform abstract mathematical theories into impactful, real-world applications. My focus now is on cutting-edge areas such as Artificial Intelligence, Neural Networks, Computer Vision, and Reinforcement Learning - fields where I can blend my analytical skills with creative problem-solving to innovate and advance the boundaries of technology.

Data Science for me is more than a profession; it's a canvas where I paint with numbers and algorithms, creating solutions that matter.