CVAT REST API to upload files

Question

I am able to create a Task in an existing project in CVAT using the below code, but I am unable to upload files, even if I try to reference this link here: https://github.com/opencv/cvat/issues/4704

Any advice would be greatly appreciated please!

import requests

# to create project (this works)
link = 'http://xx.x.xxx.xx:8080/api/tasks'
d = {
"name": "ABC",
"project_id": 3
}

Header = {
"X-Organization":"ABC_labelling"
}

r = requests.post(link, data=d, headers=Header, auth=('username','pw'))
r.json

# to upload file (doesn't work)
task_id = 8
link = 'http://xx.x.xxx.xx:8080/api/tasks/{task_id}/data/'
files = [
'/Users/username/Documents/images/abc001.tif', 
'/Users/username/Documents/images/abc002.tif'
]

Header = {
"X-Organization":"ABC_labelling",
"Upload-Start":"true",
"Upload-Finish":"true"
}

d = {
"image-quality": 70,
"server_files": files
}

r = requests.post(link, data=d, headers=Header, auth=('username','pw'))

Answer 1

While it's sometimes a bit buggy for me. Voxel51 is by far the easiest way to achieve this.

# A unique identifier for this run
import fiftyone as fo 

# Just the CVAT password and username
assert os.environ['FIFTYONE_CVAT_URL'], 'Set FIFTYONE_CVAT_URL env var'
assert os.environ['FIFTYONE_CVAT_PASSWORD'], 'Set FIFTYONE_CVAT_PASSWORD env var'

dataset = fo.Dataset.from_dir(
    dataset_dir=<coco-dir-path>,
    dataset_type=fo.types.COCODetectionDataset,
)
dataset.name = "example_task_name"

anno_key = 'something_voxel51_uses'

dataset.annotate(
    anno_key,
    label_field="ground_truth",
    attributes=["iscrowd"],
    launch_editor=True,
    classes=['car','truck','person'],
)

https://docs.voxel51.com/api/fiftyone.core.collections.html#fiftyone.core.collections.SampleCollection.annotate

I may be missing some detail but the CVAT Api is a pain to use, while the voxel51 design is much simpler.