A5103091268- Multimodal Machine Learning Applications
- Human Pose and Action Recognition
- Domain Adaptation and Few-Shot Learning
- Advanced Image and Video Retrieval Techniques
- Natural Language Processing Techniques
- Topic Modeling
- Video Surveillance and Tracking Methods
- Face recognition and analysis
- Video Analysis and Summarization
- Advanced Text Analysis Techniques
Information Technology University
2023-2024
University of Engineering and Technology Lahore
2018-2021
Laboratoire d'Informatique de Paris-Nord
2019
Summarization generates a brief and concise summary which portrays the main idea of source text. There are two forms summarization: abstractive extractive. Extractive summarization chooses important sentences from text to form whereas paraphrase using advanced nearer-to human explanation by adding novel words or phrases. For annotator, producing document is time consuming expensive because it requires going through long composing short summary. An automatic feature-rich model for proposed...
After the September 11 attacks, security and surveillance measures have changed across globe. Now, cameras are installed almost everywhere to monitor video footage. Though quite handy, these produce videos in a massive size volume. The major challenge faced by agencies is effort of analyzing data collected generated daily. Problems related twofold: (1) understanding contents streams, (2) conversion condensed formats, such as textual interpretations summaries, save storage space. In this...
In recent years, deep learning approaches have gained great attention due to their superior performance and the availability of high speed computing resources. These are also extended towards real time processing multimedia content exploiting its spatial temporal structure. this paper, we propose a learning-based video description framework which first extracts visual features from frames using convolutional neural networks (CNN) then pass derived representations into long-short term...
In this research study, we have proposed an appearance-based classifier to classify human corresponding their attire and clothing. Personality re-identification proposes address the issue of keeping track persons. Our methodology presents advance work on both CUHK03 CUHK01 dataset, bypass over-fitting problem by artificially enlarging dataset using label-preserving transformations. The model is fine-tunned a small amount target which achieved results equal state-of-the-art. Deep...