mmdetection/tests/test_visualization/test_local_visualizer.py

import os
from unittest import TestCase

import cv2
import numpy as np
import torch
from mmengine.structures import InstanceData, PixelData

from mmdet.evaluation import INSTANCE_OFFSET
from mmdet.structures import DetDataSample
from mmdet.visualization import DetLocalVisualizer


def _rand_bboxes(num_boxes, h, w):
    cx, cy, bw, bh = torch.rand(num_boxes, 4).T

    tl_x = ((cx * w) - (w * bw / 2)).clamp(0, w)
    tl_y = ((cy * h) - (h * bh / 2)).clamp(0, h)
    br_x = ((cx * w) + (w * bw / 2)).clamp(0, w)
    br_y = ((cy * h) + (h * bh / 2)).clamp(0, h)

    bboxes = torch.stack([tl_x, tl_y, br_x, br_y], dim=0).T
    return bboxes


def _create_panoptic_data(num_boxes, h, w):
    sem_seg = np.zeros((h, w), dtype=np.int64) + 2
    bboxes = _rand_bboxes(num_boxes, h, w).int()
    labels = torch.randint(2, (num_boxes, ))
    for i in range(num_boxes):
        x, y, w, h = bboxes[i]
        sem_seg[y:y + h, x:x + w] = (i + 1) * INSTANCE_OFFSET + labels[i]

    return sem_seg[None]


class TestDetLocalVisualizer(TestCase):

    def test_add_datasample(self):
        h = 12
        w = 10
        num_class = 3
        num_bboxes = 5
        out_file = 'out_file.jpg'

        image = np.random.randint(0, 256, size=(h, w, 3)).astype('uint8')

        # test gt_instances
        gt_instances = InstanceData()
        gt_instances.bboxes = _rand_bboxes(num_bboxes, h, w)
        gt_instances.labels = torch.randint(0, num_class, (num_bboxes, ))
        det_data_sample = DetDataSample()
        det_data_sample.gt_instances = gt_instances

        det_local_visualizer = DetLocalVisualizer()
        det_local_visualizer.add_datasample(
            'image', image, det_data_sample, draw_pred=False)

        # test out_file
        det_local_visualizer.add_datasample(
            'image',
            image,
            det_data_sample,
            draw_pred=False,
            out_file=out_file)
        assert os.path.exists(out_file)
        drawn_img = cv2.imread(out_file)
        assert drawn_img.shape == (h, w, 3)
        os.remove(out_file)

        # test gt_instances and pred_instances
        pred_instances = InstanceData()
        pred_instances.bboxes = _rand_bboxes(num_bboxes, h, w)
        pred_instances.labels = torch.randint(0, num_class, (num_bboxes, ))
        pred_instances.scores = torch.rand((num_bboxes, ))
        det_data_sample.pred_instances = pred_instances

        det_local_visualizer.add_datasample(
            'image', image, det_data_sample, out_file=out_file)
        self._assert_image_and_shape(out_file, (h, w * 2, 3))

        det_local_visualizer.add_datasample(
            'image', image, det_data_sample, draw_gt=False, out_file=out_file)
        self._assert_image_and_shape(out_file, (h, w, 3))

        det_local_visualizer.add_datasample(
            'image',
            image,
            det_data_sample,
            draw_pred=False,
            out_file=out_file)
        self._assert_image_and_shape(out_file, (h, w, 3))

        # test gt_panoptic_seg and pred_panoptic_seg
        det_local_visualizer.dataset_meta = dict(classes=('1', '2'))
        gt_sem_seg = _create_panoptic_data(num_bboxes, h, w)
        panoptic_seg = PixelData(sem_seg=gt_sem_seg)

        det_data_sample = DetDataSample()
        det_data_sample.gt_panoptic_seg = panoptic_seg

        pred_sem_seg = _create_panoptic_data(num_bboxes, h, w)
        panoptic_seg = PixelData(sem_seg=pred_sem_seg)
        det_data_sample.pred_panoptic_seg = panoptic_seg

        det_local_visualizer.add_datasample(
            'image', image, det_data_sample, out_file=out_file)
        self._assert_image_and_shape(out_file, (h, w * 2, 3))

        # class information must be provided
        det_local_visualizer.dataset_meta = {}
        with self.assertRaises(AssertionError):
            det_local_visualizer.add_datasample(
                'image', image, det_data_sample, out_file=out_file)

    def _assert_image_and_shape(self, out_file, out_shape):
        assert os.path.exists(out_file)
        drawn_img = cv2.imread(out_file)
        assert drawn_img.shape == out_shape
        os.remove(out_file)