iTAG writes annotation results to a .manifest file. This page describes the top-level JSON structure of that file and how each field maps to the labeling, review, and acceptance stages.
Each .manifest file is a JSON object with the following top-level structure. The three stage keys are dynamic: replace xxxxx(Job ID) with your actual job ID.
{
"data": {
"source": "oss://********/pics/fruit/apple-1.jpg"
},
"label-xxxxx(Job ID)": {
"results": [{...}, {...}, {...}]
// Lists the items in the order they appear in the labeling stage. For the data structure of each element, see the description below.
},
"label-xxxxx(Job ID)-check": {
"results": [{...}, {...}, {...}]
// Lists the items in the order they appear in the review stage. For the data structure of each element, see the description below.
},
"label-xxxxx(Job ID)-verify": {
"results": [{...}, {...}, {...}]
// Lists the items in the order they appear in the acceptance stage. For the data structure of each element, see the description below.
}
}The following list describes each top-level field:
data: The object to be annotated. The
sourcesub-field contains the OSS path to the input file.label-xxxxx(Job ID): Annotation results from the labeling stage. The key name includes your job ID.
label-xxxxx(Job ID)-check: Annotation results from the review stage.
label-xxxxx(Job ID)-verify: Annotation results from the acceptance stage.
-
results: The annotation payload for a given stage, as an ordered array of items. The schema of each item depends on the template type used in the job. See the following topics for template-specific schemas: