Computer Vision Projects

Prediction Of Life Qualities

• The project aims to predict the life quality of a city by using numerical data.
• Project blogged every week on medium.
• 85 features(Hospital, school, park numbers per capita, etc.) are extracted from JSON files.
• Classification and regression models were developed.
• SVM, Decision Tree, and Random forest algorithms were used.

WAD Video segmentation

• It is a segmentation task containing cars, roads, sky and etc.
• U-net model is used.
• 30GB(20 training-10test) data was used for training and evaluation.
• Keras ve tensorflow is used.

Graduation project

• It is a UAV project that includes Remote Pilot Station(RPS), a Fly control system(FCS), and object tracking systems.
• UAV could not complete because of a fly accident.
• Data communication between Raspberry pi-RPS
• Tracking Algorithms(MIL, KCF, TLD, MEDIANFLOW, GOTURN, MOSSE,CSRT) research and application
• Data production for car detection
• Tests on raspberry pi.

Intern/Plate Recognition System

• Project completed in 3 weeks.
• Plate Detection&Recognition system developed.
• Data preparation and OCR system development is done because of tesseract is not stable enough.
• 1200 images labeled and trained.
• Rcnn and yolo models were used.
• Darknet, Keras, and OpenCV were used.

Teknofest AI 2019

• We qualified for the finals for the first 50 teams and placed 15th.
• RFCN, RCNN, SSD, YOLO, and FPN models research and comparison were done.
• RFCN, RCNN, and Yolo models were trained and evaluated.
• Tensorflow, OpenCV, sklearn, and darknet were used.

Captha Resolve Algorithm

• Random symbols are placed randomly placed on the random images with random colors and ratios. The task is to find places of symbols given order.
• Template matching, surf matching, object detection, and classification algorithms were used to solve problems.
• The above algorithms were used together by a voting mechanism.
• Efficientnet was used because of size and speed concerns.