Computer Science, Mathematics & Robotics

According to the New York Times, the final death toll due to pill double dose in 2016 was just over 56,000 people. When I heard about this, I was shocked. For my science fair project, I was going to solve this problem, mainly for the elderly. I would make a system that can detect weight loss in a pill bottle. When someone does take the pill, an Arduino computer based will plot the weight change on a graph on a display located near the system.

I had two ideas for the pill scale. One measured with pressure and the other one used a load cell. To make a weight scale, you can use a balloon or a load cell. Using a filled balloon, you put the pill bottle over it in a secure container. Next you connect a pressure gauge to the balloon and also connect it to the Arduino. Every time you take out a pill, the balloon will expand. This will be read by the pressure gauge that will send the message to the Arduino. Finally, the Arduino will plot the weight change on the display. You can also make a weight scale using a load cell. A load cell is a transducer with a hole in the middle that bends slightly when something is put on it. There is a strain gage near the hole. It is used to create an electric signal whose magnitude is directly correspondent to the force measured. First, you 3D print a circle with a hole in the middle in a way that a pill bottle can fit in it. Second, you screw it into one side of the strain gage. You place the pill bottle in the 3D printed holder. Next, you connect the load cell to the HX711 which measures the signal from the strain gage and sends it to the Arduino. Whenever a pill is removed from the pill bottle, the load cell will detect it. The HX711 will pass it onto the Arduino. Finally the Arduino will plot the weight change against time on a graph. By looking at the graph you can figure out if you took a pill or not.

I figured that using a balloon wouldn’t be the smartest choice for a weight scale. Since change in temperature and other factors could affect the size of the balloon, this would result in inaccurate results and would make the problem of pill double dose worse. With a load cell, temperature can’t affect it and neither can other factors, I chose this method. This is why the load cell would be more effective and not make the world wide issue worse.

For my project, I will evaluate how difficult different types of passwords are to crack. Many people fear that they may be hacked or have their identity stolen. I am hypothesizing a Caesar cipher encrypted in a password will make the it harder to crack. If my testing is correct, than it will keep people safe from hackers. First I will write a simple password-guessing program to try to guess 20 sample passwords. In order to do this, I will use a computer, and a download of Python 3. Python 3 is a program that will allow me code a password strengthener. Then, I am going to write a Caesar Cipher that shifts by 3. I will test to see if cipher codes will strengthen the password and possibly give hackers trouble with breaking your password. In the end, I found that my prediction was correct. The ciphered version of the passwords took slightly longer to crack than the normal versions. The ciphered passwords were harder by nanoseconds and milliseconds. Either way, the ciphered passwords took longer to crack, meaning that it gives hackers a harder time. In conclusion, my hypothesis tested that passwords will be harder to hack if they are encrypted into a Caesar cipher.

To cater to the needs of thousands of visually impaired individuals, I developed a cheaper and improved smart cane. This device uses ultrasonic components to detect obstacles in the direct path of a blind person and alerts the user with sounds and vibrating motion. The smart cane has built in sensors to detect wet ground and flashing LED lights to attract nearby people. An Arduino-based microcontroller is programmed to control various components that are built into the smart cane. Current solutions for the blind range from guide dogs to the white cane. Intelligent white canes do exist but are priced at above $5000. Throughout the building and coding process, I have been looking at ways to improve the user experience. Several prototypes were built to improve the usability and to reduce the weight of the cane. I also included a rechargeable battery and a USB interface to charge the device. Panic buttons and on-off switches are distinct and located conveniently. The smart cane automatically illuminates in the dark to alert nearby people. This cane is also very similar to the original white cane, with a black grip and the white center rod. I ensured that the price is under $50, making it a viable alternative to the basic white cane with key features from the commercially available smart cane. By introducing this device, I hope to eliminate the pricing constraint, and improve the lifestyle of blind people who cannot afford the $5000 cane.

The number of mitotic events in breast histology is an important indicator for analyzing breast cancer prognosis. Determining the number of mitotic events is typically done manually by counting each instance of mitosis in a histology image, which is tedious and subjective. In the advent of promising machine learning and computer vision models, we investigate the optimization of mitosis classification using deep convolutional neural networks (CNN) through transfer learning by fine-tuning CNN layers. Specifically, we look at how fine-tuning AlexNet, VGG16, and VGG19 models differently affects their respective classification accuracies. We found that fine-tuning the final fully-connected layer of the VGG16 CNN yields the highest classification accuracy. Using a reserved image test set, this model is able to yield a classification accuracy of 91.33%. We then used our optimized classification model to implement a selection-search inspired scanning algorithm to localize mitotic events in histology images, which is able to pinpoint general regions of mitosis in a histology slide. The methodology used in this research can be easily generalized and applied to other medical imaging tasks.

The Image Analysis (Computer Vision) is a very important Artificial Intelligence technology for our daily life. Currently three famous technology companies, Google (Google Cloud Vision Service v1), Amazon (Amazon Rekognition Services v1) and Microsoft (Cognitive Services Vision and Face v1) are providing their Image Analysis services to public. All of them claim to support Label Detect, Face Detect and OCR (Optical Character Recognition). This project endeavors to find out which company is the best in image analysis in 3 different categories. The hypothesis is that Google's AI should be the best in all three categories because Google's DeepMind AI has shown itself to be very capable in playing Go through its impressive data compilation and analyzation. The experiment used common computer software utility to obtain the image analysis services of Google, Amazon and Microsoft to analyze the same set of images. The results of these requests were reviewed and scored based on correctness, accuracy and comprehensiveness. The scores were logged into Google spreadsheet for summary and comparison. The experiment results did not support the initial hypothesis that the Google AI is the best only in label detection. Instead, Amazon is overall the best because it is leading in both face detection and OCR (text detection). The findings will help people choose the best image analysis for label detection, facial identification and image to text conversion.

For our venture in the Alameda Science Fair, we have designed and engineered a pet door. But there is more to this apparatus than meets the eye. Throughout the process of building our automated genie, we had to solve complex coding and mechanical problems. We learnt how to resolve the complexities, both in code and mechanics and come up with an extraordinary design. But in the end were were able to reach our main goal of designing the perfect pet door. We had to make sure every piece of our project interacted with the other in perfect harmony. We needed to make a pet door that enables any pet to come and go, in and out of their house with ease and without being followed by any stray pets or wild animals. We also wanted to make the door so no matter how unkempt or rough the pet using the door was, the door would never lose its sturdy frame. Our team decided that we would use a cardboard box for our prototype because at the time cardboard was the most abundant and strongest material available to us. Using an old cardboard box, we took precise and perfect measurements to make an entrance and exit in the door. We then made a cardboard platform on the bottom of the door that was glued to the bottoms of the exits. On the top middle of the cardboard box are a set of magnets that are supposed to keep the door in place. To make the door we used old cereal boxes and magnets bought to fit under the door and on the door frame. Besides the mechanical aspect of the project, we also worked on integrating the Raspberry Pi and arduino applications into the pet door. This ingenious device also includes a camera which takes pictures of the pet as it enters and exits through the door. In addition to a camera, we have been able to integrate a locking system for the door that used a servo motor as a guider to help lock the door. Our team used languages including python to code the Raspberry Pi and C/C++ to code the arduino. In addition, the Raspberry Pi and Arduino work together via serial usb connection. The Raspberry broadcasts a webpage with Node-Red. The website calls on a MySQL database from which it checks if the rfid tag is approved for that specific door. In conclusion, we have executed an idea which gives pet owners more clarity into the safety of their pets while they enter and exit their homes.

This is a simple project that is meant to make studying wildlife easier by scientists and animal lovers. The wildlife camera has an abundance of features allowing it to be used in a variety of environments.These features include an infrared camera to take thermal images, a microphone so it can record sound, a motion detector to identify if something is moving, a camera to take pictures, time lapses, and videos. Within this, this camera is also self-operated, internet capable, programmable, and wireless. All of the images and videos will be stored into a drive which can be sent wirelessly to the receiver or can be regularly checked upon by owners. It is also very cheap compared to other cameras and offers a wider range of features. The camera is cost effective and reaches out to a variety of people. For example, classes that focus around studying wildlife can use this object because it is cheap and the money barrier doesn’t prevent them from using it. Enthusiasts who spend hours waiting to capture the right moment to view a natural event, don’t have to anymore. They have to wait since they have the camera to rely on. It helps them capture the beauty of nature.