A web-based exercise annotation tool originally created for GymCam.
Run ng serve for a dev server. Navigate to http://localhost:4200/. The app will automatically reload if you change any of the source files
Run ng build to build the project. The build artifacts will be stored in the dist/ directory. Use the -prod flag for a production build
While this tool is created for labeling gym activities and counting the repetitions, it can be used more generically for any labelling tasks of videos recorded from a stationary camera.
To start, load a video supported by your browser. One caveat is that as of now, the tool does not detect video FPS automatically and assumes a hard-coded value in /src/app/video-metadata.ts.
After the video is loaded, you can continue from a previously saved annotation file or create a new one. When an annotation file is loaded/created, you will see a paperclip symbol next to the time code display.
Whenever you have made some changes to loaded annotation file, you will see a \* sign next to the paperclip symbol. You can save the annotation file in JSON format. The saved file contains annotation file version (currently 3), timestamp, video metadata, and block infos.
Pause at the frame which the annotated action begins, and draw the annotation block on top of the video. You need to give the block a name. The name does not have to be unique, and the tool will suggest a name that you can use which is not guaranteed to be unique.
At any point of the video, you can click on a block to add a text annotation that is associated with that particular frame. When you mark the end frame of a block, you will be prompt to add a summary, a special type of text annotation. In GymCam, it is used for the number of repetitions of some exercise, but you can use it for whatever string that fits your needs.
When a block is first created, it has a end frame of the last frame of the video. You can edit the endpoints of a block (both start and end frame).
If needed, you can add a class name to each annotation block. Just click on the block in the list, and enter a name. All used names will be stored locally as autocomplete candidates. You can edit the default types in /src/app/tagging-tool.component.ts.