Embedded Files
PROJECT TITLE:
Development of Segmentation-based Unified Deep Tracking Models for AIVSS System using Artificial Intelligence Techniques
DESCRIPTION:
An innovative intelligent monitoring system has been developed for advanced intelligence video surveillance systems (AIVSS) in real-time dynamic environments, seamlessly integrating video object segmentation and object tracking systems. Initially, a semi-supervised video object segmentation framework was created to classify and identify target objects by partitioning images into distinct regions, each corresponding to a meaningful object or structure. Subsequently, the visual object tracking framework was developed within the intelligent monitoring system to track specific objects in the given videos. This framework extensively researched various studies, exploring spatial-temporal, context-aware, key-filter, sparse spatial, sparse temporal, spatial variance, and more. Ultimately, these two techniques were combined to enhance object region identification and ensure accurate continuous tracking within the AIVSS framework.
OVERALL PROJECT FRAMEWORK
SKILLS:
Matlab
Visual Studio
CUDA GPU
Python
Tensorflow
PyTorch
Anaconda
KEY ROLE:
Specialized in machine learning techniques (Lasso, Ridge Regression, etc.).
Specialized in video object segmentation (VOS) (Semantic, Instance).
Proficient in deep feature extraction methods (VGG16, VGG19, ResNet50, etc.).
Proficient in developing deep-tracking frameworks using Siamese networks.
Highly skilled in handling one-shot learning, contour, and graph cut methods.
Extensive knowledge acquired in Python, Tensorflow, PyTorch, and Anaconda.
Proficient in software tools such as MATLAB, Visual Studio, and CUDA GPU.
Achieved optimal results in segmentation-based object tracking.
Achieved best tracking results and comparative analysis with the benchmarks.
PROJECT OUTCOMES
VIDEO OBJECT SEGMENTATION (Instance Segmentation)
VIDEO OBJECT TRACKING (Object Tracking)
ADVANCED INTELLIGENCE VIDEO SURVEILLANCE SYSTEM (AIVSS)
(Segmentation-based Object Tracking)
Google Sites
Report abuse