datasets/build_multi.py

import json
import os
from PIL import Image
import numpy as np
from pycocotools.mask import encode, decode, frPyObjects
from tqdm import tqdm
import copy

if __name__ == '__main__':
    
    new_img_id = 0
    nerf_dataset = []
    first_frame_annotation_relpath = "teacup/Annotations/000334.png"
    first_frame_img_relpath = "teacup/JPEGImages/000334.png"
    first_frame_annotation_path = "/home/yuqian_fu/Projects/sam2/teacup/Annotations/000334.png"
    first_frame_annotation_img = Image.open(first_frame_annotation_path)
    first_frame_annotation = np.array(first_frame_annotation_img)
    height, width = first_frame_annotation.shape
    unique_instances = np.unique(first_frame_annotation)
    unique_instances = unique_instances[unique_instances != 0]
    print("unique_instances:", unique_instances)
    coco_format_annotations = []
    # for semi-supervised VOS, we use first frame's GT for input
    for instance_value in unique_instances:
        binary_mask = (first_frame_annotation == instance_value).astype(np.uint8)
        segmentation = encode(np.asfortranarray(binary_mask))
        segmentation = {
            'counts': segmentation['counts'].decode('ascii'),
            'size': segmentation['size'],
        }
        area = binary_mask.sum().astype(float)
        coco_format_annotations.append(
            {
                'segmentation': segmentation,
                'area': area,
                'category_id': instance_value.astype(float),
            }
        )


    sample_img_relpath = "teacup/JPEGImages/000345.png"
    image_info = {
        'file_name': sample_img_relpath,
        'height': height,
        'width': width,
    }

    print("coco_format", len(coco_format_annotations))
    sample_annotation_path = "/home/yuqian_fu/Projects/sam2/teacup/Annotations/000345.png"
    sample_annotation = np.array(Image.open(sample_annotation_path))
    sample_unique_instances = np.unique(sample_annotation)
    sample_unique_instances_tmp = sample_unique_instances[sample_unique_instances != 0]
    sample_unique_instances = []
    for i in unique_instances:
        if i in sample_unique_instances_tmp:
            sample_unique_instances.append(i)

    print("sample_unique_instances:",sample_unique_instances)
    anns = []
    for instance_value in sample_unique_instances:
        assert instance_value in unique_instances, 'Found new target not in the first frame'
        binary_mask = (sample_annotation == instance_value).astype(np.uint8)
        segmentation = encode(np.asfortranarray(binary_mask))
        segmentation = {
            'counts': segmentation['counts'].decode('ascii'),
            'size': segmentation['size'],
        }
        area = binary_mask.sum().astype(float)
        anns.append(
            {
                'segmentation': segmentation,
                'area': area,
                'category_id': instance_value.astype(float),
            }
        )
    first_frame_anns = copy.deepcopy(coco_format_annotations)
    if len(anns) < len(first_frame_anns):
        first_frame_anns = [ann for ann in first_frame_anns if ann['category_id'] in sample_unique_instances]
    print("anns:",len(anns))
    print("first_frame_anns",len(first_frame_anns))
    assert len(anns) == len(first_frame_anns)
    sample = {
        'image': sample_img_relpath,
        'image_info': image_info,
        'anns': anns,
        'first_frame_image': first_frame_img_relpath,
        'first_frame_anns': first_frame_anns,
        'new_img_id': new_img_id,
        'video_name': "teacup",
    }
    nerf_dataset.append(sample)
            

    save_path = "/home/yuqian_fu/Projects/sam2/predicted_mask/kitti_1.json"
    with open(save_path, 'w') as f:
        json.dump(nerf_dataset, f)
    print(f'Save at {save_path}. Total sample: {len(nerf_dataset)}')