Mmdetection mask

[DrawingProcess]

Description

See #765

List any dependencies that are required for this change.

Type of change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)

How has this change been tested, please provide a testcase or example of how you tested the change?

Try putting in a configuration without a mask and a config with a mask to see if the value fits well

Any specific deployment considerations

Nothing

Docs

There is nothing special

[CLAassistant]

All committers have signed the CLA.

[SkalskiP]

Hi @DrawingProcess 👋🏻 ! Thanks a lot for your interest in supervision. Could you provide an example that we could use to test your changes?

[DrawingProcess]

Here is a test code.

First of all, You have to ready for mmdetection Environment Then modify datapath(datset, sub(coco format)) of the below code.

test code

import random
from mmdet.apis import init_detector, inference_detector

import supervision as sv

from mmengine import Config

# without mask model
# cfg = Config.fromfile('checkpoint/rtmdet_s_8xb32-300e_coco.py')
# checkpoint_file = 'checkpoint/rtmdet_s_8xb32-300e_coco_20220905_161602-387a891e.pth'

# with mask model
cfg = Config.fromfile('checkpoint/rtmdet-ins_s_8xb32-300e_coco.py')
checkpoint_file = 'checkpoint/rtmdet-ins_s_8xb32-300e_coco_20221121_212604-fdc5d7ec.pth'

dataset = "/home/avs/dataset/train_exd/"
sub = "EVT_20230715_112856_F_trim"

ds = sv.DetectionDataset.from_coco(
    images_directory_path=dataset + sub,
    annotations_path=dataset + sub + ".json",
)

images = list(ds.images.values())

model = init_detector(cfg, checkpoint_file)

# image = random.choice(images)
with sv.ImageSink(target_dir_path=f"/home/avs/dataset/train_vis_rtmdet/{sub}", overwrite=True) as sink:
    for i, image in enumerate(images):
        result = inference_detector(model, image)
        detections = sv.Detections.from_mmdetection(result)
        print(detections)

Result

without mask model with mask model

[SkalskiP]

Hi @DrawingProcess 👋🏻 Thank you very much for your contribution! Would it be possible for you to share a Google Colab link for testing purposes?

Establishing a local testing environment could be quite time-consuming, and that time could be more effectively allocated towards developing new features or enhancing our documentation. A Google Colab link would significantly streamline our process, enabling us to integrate this pull request more swiftly.

We greatly appreciate your understanding and support in this matter.