For pipelines and training we need to attempt to use the source dataset if possible.

Notes:

• We currently have the skip transcoding option that can be used if videos are uploaded to prevent automatic transocding of videos.
• There are a bunch of checks that are done to check and see if the video file needs to be transcoded. These include issues like double frames or missing frames.

Feature:

• Option for training and Pipelines that the user can use the source video/images vs the transcoded images.
• By default I think it should default to the source video file.
• There should be a help icon indicating that the pipeline will be run on the source video unless there is something that required the file to be transcoded because of mis-alignment or some other issue.

Implementation:

• Add a metadata key to source videos where transcoding is done because of a mis-alignment or some other issue
• Default to skip transcoding instead of the option to skip transcoding on upload
• Check and see if S3 Imports have Skip Transcoding on (Checked and it is)
• Add to pipelines/training endpoint the option to utilize the transcoded or source video file.
• In the training and pipeline tasks handle this issue. Note: Not all files have a transcoded version, if the video file didn't require transcoding it won't be displayed. Use the logging output to indicate what the value should be
• Add skiptranscoding options to the desktop version as well and default to non-transcoded version,