A number of projects are available on this page. Each project must be documented thoroughly using a Jupiter notebook or a GitHub repository. Send your GitHub id to huber DOT flores AT ut DOT ee
PROJECT 1: Analyzing handgrip strength dataset: Handgrip strength is typically measured using a dynamometer. However, it is possible to envision the monitoring of handgrip strength using light sensors. In this project, a dataset that contains the handgrip strength of multiple participants is provided. The overall goal is to analyze the data, categorize handgrip strength and create a deep learning prediction model to identify individuals. ( Reserved: Elvin Mirzazada)
PROJECT 2: Ear sensing - understanding typing patterns of users on mobile screens through typing sound: Microphones are sensitive enough that can be used to capture the typing sound on the screen of a smartphone. In this project, a mobile application is given that captures sounds and gestures performed by a user on a mobile screen. The main goal of this project is to analyze the data and conduct a user study with at least 15 participants, such that it is possible to analyze whether patterns can be created with typing sounds. ( Reserved: Lukas Andrijauskas)
PROJECT 3: Review Explainability methods for Artificial Intelligence: Several explainability methods are available to understand model execution and training. The goal of this project is to review existing methods for explainability. From this review, a particular method has to be studied in detail, e.g., permutation, occlusion, through examples and uses cases. A full descriptive report is the deliverable of this project.
PROJECT 4: Analyze whether is possible to characterize different types of drinks through WiFi. In this project, an application to collect WiFi signal between two phones is provided. The goal of the project is to analyze whether a WiFi fingerprint can be established for a drink (in a glass) as the drink is located between the phones. Moreover, as part of the analysis, it is also possible to analyze the effect of mixing different drinks.
PROJECT 5: Perception of users towards autonomous vehicles invading public spaces. Unmanned autonomous vehicles (UAV)s are automating many tasks, such as grocery and packet delivery. However, UAVs need to move autonomously through public urban spaces, which can cause problems with pedestrians. In this project, a survey with pictures of UAVs invading public spaces should be carried out. The survey should consider at least the opinions of 30 participants. Conclusions on how to address the problem should be given. ( Reserved: Dmytro Zabolotnii )
PROJECT 6: Review of Federated Learning (LF) for model training. The goal of this project is to identify a list of possible attacks that can be performed in FL. Moreover, an attack over data that is use to train the model should be implemented (aka data poisoning attack). The outcome of this project is a systematic analysis, demonstrating the differences in model performance (based on accuracy) when training a FL with/without data poisoned. ( Reserved: Amey Darekar )
PROJECT 7: Identifying devices connected to the network using traffic analysis. The goal of this project is to characterize the type a device using its data transferred through the network. THe overall idea is to monitor network activity of different types of devices, e.g., Wireshark, and then analyzing the data to extract fingerprints that can be used to depict a type of device.
PROJECT 8: Review about why digital contact tracing failed in pandemic times. In this project, a taxonomy of existing work about digital contact tracing should be created. Taxonomy is not a review, but it requires reviewing a certain amount of work. A good example of a taxonomy can be found here
PROJECT 9: User perception of litter classification using sensors. Separation of litter is a fundamental process to recycle reusable materials. Unfortunately, when materials get mixed with other waste objects, it is difficult to separate them - sometimes even impossible. As a result, litter classification approaches for early separation of litter waste are required. In this project, we will conduct a user study, in which several participants dispose of waste in trash bins using traditional visual inspection approaches, and then we compare with a recommendation system that suggests to users which trash bin is the right one (using light sensors in a glove).
PROJECT 10: Nutritional value of fruits and vegetables. In our previous project, we demonstrated that it is possible to capture the decomposition of fresh produce using sensors. In this project, we want to verify whether it is possible to link the nutritional value of produce (fruits and vegetables) to their decomposition state that was captured by the sensor.