SAVE: A framework for semantic annotation of visual events

By Mun Wai Lee, Asaad Hakeem, Niels C. Haering, Song-Chun Zhu. 2008

43 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

1. Marcus Rohrbach
2. Trevor Darrell
3. Yi Yang
4. Zhe Gan
5. Mohit Bansal
6. Jeff Donahue
7. Yoshua Bengio
8. Haonan Yu

Citations in the corpus

• Video Paragraph Captioning Using Hierarchical Recurrent Neural Networks (2015, methodology): “Following their work, several succeeding approaches [26, 20, 21, 15, 3] applied similar rule-based systems to datasets with larger numbers of objects and events, in different tasks and scenarios.”
• Learning to Describe Video with Weak Supervision by Exploiting Negative Sentential Information (2015, methodology): “First, some methods craft rule-based systems in which language generation is treated as an engineering task (Kojima, Tamura, and Fukunaga 2002; Lee et al. 2008; Khan, Zhang, and Gotoh 2011a; 2011b; Hanckmann, Schutte, and Burghouts 2012; Siddharth, Barbu, and Siskind 2014).”
• Towards coherent natural language description of video streams (2011, unknown): “[10] proposed the SAVE framework for video annotations consisting of following steps; identifying objects, finding their relations, detecting events and representing them in a human readable format.”
• Translating Videos to Natural Language Using Deep Recurrent Neural Networks (2014, background): “The memory cell is at the core of the LSTM unit and it is modulated by the input, output and forget gates controlling how much knowledge is transferred at each time step. which incorporate explicitly controllable memory units, are known to be able to learn long-range temporal dependencies.”
• Human Focused Video Description (2011, background): “Lee et al. introduced a framework for semantic annotation of visual events in three steps; image parsing, event inference and language generation [11].”
• YouTube2Text: Recognizing and Describing Arbitrary Activities Using Semantic Hierarchies and Zero-Shot Recognition (2013, background): “The existing work on describing videos with sentences [12, 17, 13, 8, 14, 5] deals with constrained domains with a limited set of actions or objects, and does not exploit text .”
• Generating Natural-Language Video Descriptions Using Text-Mined Knowledge (2013, background): “The small amount of existing work on videos (Ding et al., 2012; Khan and Gotoh, 2012; Kojima et al., 2002; Lee et al., 2008; Yao and Fei-Fei, 2010) uses hand-crafted templates or rule-based systems, works in constrained domains, and does not exploit text mining.”