[torchcodec] initial version of simple video decoder #42
Conversation
facebook-github-bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D58605628 |
|
This pull request was exported from Phabricator. Differential Revision: D58605628 |
|
This pull request was exported from Phabricator. Differential Revision: D58605628 |
|
This pull request was exported from Phabricator. Differential Revision: D58605628 |
|
This pull request was exported from Phabricator. Differential Revision: D58605628 |
|
This pull request was exported from Phabricator. Differential Revision: D58605628 |
|
This pull request was exported from Phabricator. Differential Revision: D58605628 |
|
This pull request was exported from Phabricator. Differential Revision: D58605628 |
|
This pull request was exported from Phabricator. Differential Revision: D58605628 |
|
This pull request was exported from Phabricator. Differential Revision: D58605628 |
|
This pull request was exported from Phabricator. Differential Revision: D58605628 |
Summary: Pull Request resolved: #42 Initial version of `SimpleVideoDecoder`. This diff supports: 1. Creating a simple decoder from file or tensor. 2. Accessing frames through random access: `decoder[i]`. Internally, this gets a frame by index. 3. Iterating over all available frames: `for frame in decoder`. Internally, this calls the core library function to get the next frame. I did not implement slice semantics for random access yet. I think that in order to support that best we'll need to add new capabilities in the core API to support getting a batch of frames from a range. We *could* use `get_frames_by_indices`, but we would end up creating very large lists. This is a partial implementation of the design: https://fburl.com/gdoc/hxotv2by Reviewed By: NicolasHug, ahmadsharif1 Differential Revision: D58605628
|
This pull request was exported from Phabricator. Differential Revision: D58605628 |
|
This pull request has been merged in d4b0111. |
Summary:
Initial version of
SimpleVideoDecoder. This diff supports:decoder[i]. Internally, this gets a frame by index.for frame in decoder. Internally, this calls the core library function to get the next frame.I did not implement slice semantics for random access yet. I think that in order to support that best we'll need to add new capabilities in the core API to support getting a batch of frames from a range. We could use
get_frames_by_indices, but we would end up creating very large lists.This is a partial implementation of the design: https://fburl.com/gdoc/hxotv2by
Reviewed By: NicolasHug, ahmadsharif1
Differential Revision: D58605628