Labels
Label / Activity logs
Label logs serve as a comprehensive record of actions performed within the Encord platform’s Label Editor.
Viewing label logs
The following scripts show how to view label logs for:
- All data units of a Project in a specified time frame. The example specifies this time frame to be 1 week.
- A specific data unit, and made by a specific user.
In the following script, ensure that you:
- Replace <private_key_path> with the path to your private key.
- Replace <project_hash> with the hash of the Project you want to view label logs for
- If you are using the Time frame script, replace
weeks=1
with the time frame you want to view label logs for. - If you are using the User and Data unit script, replace <data_hash> with the hash of the data unit you want to view label logs for.
Label log content
Actions do not need to be submitted to appear in the log.
Logs are generated in chronological order and contain the following:
Log entry | Description | Note |
---|---|---|
log_hash | This log entry’s ID | |
user_hash | The user’s ID | |
user_email | The user’s email address | |
annotation_hash | The label instance’s ID | ”None” for actions that aren’t labels or attributes |
identifier | The label instance’s ID | ”None” for actions that aren’t labels or Attributes |
data_hash | The data unit’s ID | |
feature_hash | The ontology feature’s ID | A feature can have multiple instances. Equivalent to the featureNodeHash found in the ontology’s JSON |
action | The action this log entry refers to | Please see the actions section for full details |
label_name | The label’s name | As specified in the ontology. Attributes are not named. |
time_taken | The time taken to create the label (in milliseconds) | “None” for non-labeling related actions |
created_at | The date and time the label action made | Date and time are given in the UTC time zone for all users |
frame | The frame number the action was taken on | ”None” refers to the first frame |
branch_name | The name for each branch. | ”Main” branch always exists (non-consensus and Consensus). One branch is create for each annotator on Consensus tasks. |
graph_node_uuid | The unique ID for each node in a Workflow. |
Log actions
The table below provides descriptions of all log actions.
Action | Description |
---|---|
0 | Add a label |
1 | Edit a label |
2 | Delete a label |
3 | Start labeling |
4 | End labeling |
11 | Submit a task |
12 | Approve a label |
13 | Reject a label |
23 | Add a classification or attribute |
24 | Edit a classification or attribute |
25 | Delete a classification or attribute |
26 | Approve a classification or attribute |
27 | Reject a classification or attribute |
28 | Submit an object or a classification label for review |
29 | Submit an attribute for review |
30 | The buffering icon appeared when the video was loading |
31 | A bitrate warning was shown |
32 | The ‘Seeking’ overlay way shown when skipping to a specific frame in a video |
33 | Approve a task. |
34 | Reject a task. |
35 | Approves a Consensus label. Logged when: 1. Approving a label in Review and Refine. 2. Choosing a winning set + click Consensus reached and when in Consensus Review. A log is created for each label in the winning label set . |
36 | Removes an Approved label from the Consensus Approved label list. Logged when you delete an Approved Label in Review and Refine . |
37 | Consensus reached on labels. Logged when choosing the winning label set + clicking Consensus reached. Logs also created for each of the label branches that are in agreement. |
Use cases
Tutorial: Checking Rejection and Approval rates for a Project
Label logs provide a detailed record of actions taken during the labeling process. You can use these logs to analyze the rejection and approval rates of labels.
-
Use the recipe provided above to view label logs.
-
Filter log entries related to the actions “Approve a label” (Action 12) and “Reject a label” (Action 13).
-
Calculate Rejection and Approval Rates.
- Divide the number of approved or rejected tasks by the total number of reviewed tasks.
- To see the rate as a percentage, multiply the result by 100.
Was this page helpful?