Add video tracking and Media tab visualization#3405
Open
claforte wants to merge 1 commit into
Open
Conversation
|
|
Author
|
The above was authored by Codex. I personally tested it locally, it seems to work well. Hope this helps, Christian |
Author
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.

Summary
This adds first-class video tracking to Aim and exposes it in the run detail UI:
aim.VideoandVideossequence support formp4,m4v,gif,mov, andwebm/api/runs/search/videos/and video blob batch loadingMediatab that groups Images, Videos, and Audio under one tab while preserving legacy media routesplaysInline, lazy blob loading, object URL cleanup, and batched visible-video blob requestsVideoobject andVideossequence in the SDK referenceMotivation
Aim already supports rich media objects such as images and audio, but training workflows that generate rollout videos, reconstruction clips, simulated environments, or model behavior videos currently need to store those artifacts outside the normal sequence/query/UI flow.
This PR keeps videos aligned with Aim's existing media model while paying attention to large-object behavior:
run.track()stays cheap during training loopsImplementation Notes
SDK and storage
aim.sdk.objects.video.Videofollows Aim'sCustomObjectpattern.Video(path=...)stores metadata and the source path, then reads bytes only during_aim_encode().Video(data=...)supports byte-backed construction for tests and small in-memory videos.Video.__deepcopy__()avoids materializing path-backed bytes when async tracking copies values before enqueueing.VideosextendsMediaSequenceBaseand is registered throughSEQUENCE_TYPE_MAP.API
caption,format,fps,size,blob_uri,index).CustomObjectApiConfigfor video search and batch endpoints.videosin restricted query parameters and addsRepo.query_videos().UI
Mediatab./runs/:hash/images,/runs/:hash/videos, and/runs/:hash/audiosstill route to the relevant Media sub-tab.IntersectionObserverTests
Ran locally:
Result:
15 passedResult:
1 passedResult: passed
Result: passed with existing repository warnings outside the touched files.
Reviewer Notes
AIM_RUN_VIDEO_SYSTEM_TESTS=1because it can download or synthesize many video samples.AIM_VIDEO_SYSTEM_DATASET_URLis not set, the system test usesffmpegto generate a moving 1080p source and five downsampled variants, then logs 100 path-backed videos under async tracking.