Condensing Video Content: Deep Learning Advancements and Challenges in Video Summarization Innovations

Abstract: The exponential increase in video uploads on platforms like YouTube, exceeding 500 hours per minute, presents critical challenges in indexing, retrieval, and navigation. Existing methods heavily depend on user-generated metadata, which is often misaligned with the actual content. To address these challenges, we conducted a systematic review of video summarization techniques employing deep learning. An initial pool of 300 research articles was screened using strict quality criteria, resulting in a final selection of 44 studies. Articles were included if they focused on video summarization, employed deep learning approaches, utilized video datasets for evaluation, and were published in English or Urdu between 2019 and 2024 in peer-reviewed journals or conference proceedings. Papers were excluded if they lacked evaluations, used non-English/Urdu datasets, or were published before 2019. This review synthesizes recent advancements, highlights practical applications, and discusses relevant datasets, offering valuable insights for researchers and practitioners seeking to enhance automated video indexing and retrieval on social networking platforms. Keywords: YouTube, Evaluations, Systematic, Indexing, Platforms.