Embedded Files
Facial Expression Recognition Through ML and DL Approaches
Facial Expression Recognition (FER) is an exciting research subject that has been presented and used in a range of domains, including prosecution systems, safety, healthcare, and man-machine interactions. Because of machine learning's exceptional success, many machine learning models are being used to improve emotion detection performance. In this paper, first, recent works on emotion detection and FER in machine learning have been reviewed. Then two machine learning techniques, Support Vector Machine (SVM) and Random Forest (RF) are studied and implemented. Finally, the machine learning algorithms were compared using different performance evaluation measures, such as accuracy, prediction, and Area under Curve, to highlight their benefits and limitations. The results showed that the RF classifier had a 7% higher accuracy than the SVM classifier for the used dataset in this study. The result also demonstrated that RF requires less time to run than the SVM classifier. These results will help highlight a few observations that will assist in making an insightful decision about which algorithm to select for each application.
Federated Learning implementation with Raspberry Pi using Pytorch.
Federated Learning (FL) is a machine learning paradigm in which each device contributes to the training model by solely calculating the lean predicated on its local training data. Newly it’s turned into a hot pursuit matter because it promises several advantages in terms of data protection and scalability. Federated learning (FL) plays an important role in IoT especially smart cities, health, and the internet of drones. Issues related to privacy (authentication) and security increased, with federated learning (FL) we can reduce this issue. In this project, we implement federated edge learning (FEL) on real-time hardware devices that’d be Raspberry Pi. Raspberry-pi will act as edge nodes and each raspberry-pi (on the client side) will first train its local model and then send gradients to its main server for training a Global model (on the server side).
Face detection and recognition with YoloV3 and SSD on Raspberry Pi (Linux).
Face detection and recognition with YoloV3 and SSD on Raspberry Pi (Linux).
For my project, I implemented a face detection and recognition system using YOLOv3 and SSD on a Raspberry Pi running Linux. The goal was to create a lightweight, yet effective solution for real-time face detection in constrained environments, leveraging the computational efficiency of the Raspberry Pi. YOLOv3 was employed for its robust detection capabilities, while SSD was used for faster processing. The system was optimized to handle multiple faces simultaneously, ensuring reliable recognition even with limited hardware resources. This project demonstrated the feasibility of deploying advanced computer vision algorithms on low-power devices for practical applications.
Plastic detection with Yolov4 and Single Shot Detector (SSD).
Plastic detection with Yolov4 and Single Shot Detector (SSD).
For my project, I developed a plastic detection system using YOLOv4 and Single Shot Detector (SSD), both trained on a custom dataset. The custom dataset was carefully curated to include various types of plastic waste in different environments, ensuring the model could effectively detect and classify plastic objects in diverse scenarios. YOLOv4 was chosen for its superior detection accuracy, while SSD provided faster processing times. The combination of these two models allowed for a balanced approach, delivering both speed and precision in detecting plastic waste, making it suitable for real-time environmental monitoring applications.
License Plate-based Security system (Image processing with Machine learning).
License Plate-based Security system (Image processing with Machine learning).
In this project, we employed advanced image processing techniques to accurately segment the characters on license plates. Morphological operations, such as dilation and erosion, were used to enhance the contrast between the characters and the background, effectively isolating the individual numbers and letters on the plate. After successfully segmenting the characters, we applied machine learning algorithms, specifically Support Vector Machines (SVM) and K-nearest neighbors (KNN), to classify and recognize the segmented characters. SVM was utilized for its ability to handle high-dimensional data and perform well with limited training data, while KNN was used for its simplicity and effectiveness in pattern recognition tasks.
NVIDIA Jetson Nano and Orion Nano-based UAV-detection system.
NVIDIA Jetson Nano and Orion Nano-based UAV-detection system.
In my project involving an NVIDIA Jetson Nano and Orion Nano-based UAV detection system, I deployed YOLOv3 models on these platforms to effectively identify and track UAVs in real-time. The project focused on optimizing the YOLOv3 architecture to run efficiently on the Jetson Nano's GPU and the Orion Nano's edge computing capabilities. By fine-tuning the model and leveraging hardware acceleration, the system achieved a balance between high accuracy and low latency, making it suitable for real-world UAV detection scenarios in various environments.
Smart Irrigation System (Embedded Systems).
Smart Irrigation System (Embedded Systems).
In the Smart Irrigation System project, I utilized embedded systems to automate the irrigation process. A soil moisture sensor was deployed to continuously monitor the moisture levels in the soil. When the sensor detected low moisture levels, the system would automatically activate the tubewell to water the fields. Conversely, when the moisture level was adequate, the system would turn the tubewell off, ensuring efficient water usage and optimal soil conditions for crop growth. This automation helps conserve water and reduces the need for manual intervention.
Mm wave Radar base localization and mapping (SLAM) with Raspberry Pi and Python.
Mm wave Radar base localization and mapping (SLAM) with Raspberry Pi and Python.
In the "mmWave Radar-based Localization and Mapping (SLAM)" project, we utilized a mmWave sensor attached to a small robot to perform simultaneous localization and mapping (SLAM). The system was implemented using Raspberry Pi and Python, allowing the robot to navigate and map its environment autonomously. The mmWave sensor provided precise distance measurements, which were processed to generate a detailed map of the surroundings, enabling the robot to localize itself within the mapped area. The integration of SLAM algorithms ensured real-time processing and accurate mapping, making it suitable for various autonomous navigation applications.
Page updated
Google Sites
Report abuse