dc.contributor.author |
Singanporia, Kushal |
|
dc.date.accessioned |
2022-02-08T06:04:04Z |
|
dc.date.available |
2022-02-08T06:04:04Z |
|
dc.date.issued |
2019-07 |
|
dc.identifier.citation |
21p. |
en_US |
dc.identifier.uri |
http://hdl.handle.net/10263/7273 |
|
dc.description |
Dissertation under the supervision of Dr. Kumar Sankar Ray |
en_US |
dc.description.abstract |
Prior introduction of neural nets to domain of computer vision, action
recognition requires specific domain knowledge. Still domain knowledge is
useful in action recognition but with availability of huge data and neural
nets, data-driven feature learning methods have emerged as an alternative.
Recent trends in action recognition uses LSTM and its various modifications,
as LSTM have memory retaining capability which other architectures lake.
In this work we performed action recognition on different tennis strokes.
Our work relay on architecture proposed By Husain, Dellen, and Torras,
2016. Architecture is comprised of various modified VGG-nets connected in
parallel. As it doesn’t include LSTM, which makes it different than other
works. |
en_US |
dc.language.iso |
en |
en_US |
dc.publisher |
Indian Statistical Institute,Kolkata |
en_US |
dc.relation.ispartofseries |
Dissertation;;2019-24 |
|
dc.subject |
Deep Learning |
en_US |
dc.subject |
VGG16 |
en_US |
dc.title |
Recognition of Strokes in Tennis Videos Using Deep Learning |
en_US |
dc.type |
Other |
en_US |