HumanSignal / label-studio

Label Studio is a multi-type data labeling and annotation tool with standardized output format
https://labelstud.io
Apache License 2.0
18.24k stars 2.29k forks source link

Single Channel Labeling With Multi-Channel Display #3893

Open WilliamJudge94 opened 1 year ago

WilliamJudge94 commented 1 year ago

Feature I would love to be able to import a dataset that has multiple channels and can give individual labels to each channel at a time.

Use Case For battery testing, there are usually multiple batteries, all performing the same charge and discharge protocol simultaneously. However, they may not be connected to one another/have no real relation to one another. The labeling would aim to identify and tag anomalous events within each battery cycling profile. Since all profiles have almost identical shapes and are at the same time stamps, loading each battery into a different channel for labeling throughput would be beneficial. Yet, as stated above, these batteries have no real anomaly relationship to one another. Labeling just a single channel with an anomalous event would be great to increase our data throughput.

Solution Currently, when a user has a time-series labeling task, a given label is placed for all channels within the time series. In addition, I would love to be able to label only a single channel at a time without propagating to the other channels.

Alternatives We could set them as separate tasks, but this 1) would decrease the labeling throughput and 2) The upcoming feature of a multi-channel plot would be of immense help since these batteries should have an almost identical profile. Any deviation would easily be seen in this multi-channel plot and could not be achieved if placed into separate tasks.

makseq commented 1 year ago

Duplicate: https://github.com/heartexlabs/label-studio/issues/2373

arnabid commented 1 year ago

+1 on this feature request. Sometimes certain features are more notable in a subset of the channels and the ability to label channels individually would be very helpful.