Evaluation segmentation API not work?

[open-mmlab]

Evaluation segmentation API not work?

[hellock]

Adding --eval bbox segm will evaluate both bbox AP and mask AP.

[Minotaur-CN]

Thanks for your quick reply!
I used to retrain using "/configs/mask_rcnn_r50_fpn_1x.py"," ./work_dirs/mask_rcnn_r50_fpn_1x/", and
"--eval bbox segm --show"
but only show the detection result, no segmentation result.

[hellock]

I see, you mean the visualization, not evaluation. The eval argument is for mAP evaluation, and the visualization method is implemented here.

Currently we only visualize the bbox, and will further add mask visualization. You can temporally enable it by replacing the original show_result() with the following method.

def show_result(self,
                    data,
                    result,
                    img_norm_cfg,
                    dataset='coco',
                    score_thr=0.3):
        assert isinstance(result, tuple)
        bbox_result, segm_result = result

        img_tensor = data['img'][0]
        img_metas = data['img_meta'][0].data[0]
        imgs = tensor2imgs(img_tensor, **img_norm_cfg)
        assert len(imgs) == len(img_metas)

        if isinstance(dataset, str):
            class_names = get_classes(dataset)
        elif isinstance(dataset, list):
            class_names = dataset
        else:
            raise TypeError('dataset must be a valid dataset name or a list'
                            ' of class names, not {}'.format(type(dataset)))

        for img, img_meta in zip(imgs, img_metas):
            h, w, _ = img_meta['img_shape']
            img_show = img[:h, :w, :]
            labels = [
                np.full(bbox.shape[0], i, dtype=np.int32)
                for i, bbox in enumerate(bbox_result)
            ]
            labels = np.concatenate(labels)
            bboxes = np.vstack(bbox_result)
            segms = mmcv.concat_list(segm_result)

            inds = np.where(bboxes[:, -1] > score_thr)[0]
            for i in inds:
                color_mask = np.random.randint(0, 256, (1, 3), dtype=np.uint8)
                mask = maskUtils.decode(segms[i]).astype(np.bool)
                img_show[mask] = img_show[mask] * 0.5 + color_mask * 0.5

            mmcv.imshow_det_bboxes(
                img_show,
                bboxes,
                labels,
                class_names=class_names,
                score_thr=score_thr)

[Minotaur-CN]

Thanks.
some error happens.

1. class 'tensor2imgs' not defined,
   from mmdet.core import tensor2imgs
2. class 'get_classes' is not defined
   from mmdet.core import get_classes
3. import np
4. maskUtils.decode this can not find in your repo. Please update

Thanks！

[hellock]

Oh I forgot to paste the import-statements.

import numpy as np
import pycocotools.mask as maskUtils
from mmdet.core import tensor2imgs, get_classes

[Minotaur-CN]

Excellent, I can show the segmentation result.
Thanks!

[Curry1201]

Hi @Minotaur-CN , I follow the answer from Dr. Chen Kai above. The "Mask_rcnn.py" under mmdetection/mmdet/models/detectors has been modified, and then run the command "python3 tools/test.py configs/mask_rcnn_r50_fpn_1x.py checkpoints/mask_rcnn/mask_rcnn_r50_fpn_1x_20181010-069fa190.pth --show" , why is it still only bbox, what am I doing wrong?

[hellock]

@Curry1201 You can try the latest version.