fix: update to export RLE mask with the shape of (width, height) order

[ktro2828]

Description

This PR fixes exported RLE mask shape from (height, width) order to (width, height) order.

How to review

How to test

test data

deepen

Polygon

Dataset: https://drive.google.com/file/d/117MTcixJsEpnTg2DGXSC_X_y4SB7li4Z/view?usp=drive_link

• Before

{
        "token": "a2db6431a515d5c7e8463070cdd2db35",
        "sample_data_token": "a1d3257e11ec9d4a587ea7053b33f1c1",
        "instance_token": "e36688fcefc953126aa80ea19f690d7b",
        "category_token": "3852e92a0d398ff0a9bf5bbc28570a79",
        "attribute_tokens": [],
        "bbox": [
            0,
            389.66209596195347,
            1920,
            675.4739482400669
        ],
        "mask": {
            "size": [
                1280,
                1920
            ],
            "counts": "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"
        },
        "automatic_annotation": false
    },

• After

{
        "token": "f129c5f4392e1bc4be12681b35632559",
        "sample_data_token": "a1d3257e11ec9d4a587ea7053b33f1c1",
        "instance_token": "0e37be3615ab4966d8dc7c8fcad4940b",
        "category_token": "64ec257a489eab27fb403c6b908f2fdf",
        "attribute_tokens": [],
        "bbox": [
            0,
            389.66209596195347,
            1920,
            675.4739482400669
        ],
        "mask": {
            "size": [
                1920,
                1280
            ],
            "counts": "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"
        },
        "automatic_annotation": false
    },

Painting

Dataset: https://drive.google.com/file/d/1G6N4J6aohRytI2v9Q72LapKZwsKdyqDl/view?usp=drive_link

• Before

{
        "token": "d8c4ba67e59f5f8155ca191cfe72b21b",
        "sample_data_token": "02382f82a702f04cdafa69d9eac0deed",
        "instance_token": "f64b1c8f34abfae0cf8cf4e803675b0c",
        "category_token": "f2c199501add240250f801b5f651f6be",
        "attribute_tokens": [],
        "bbox": [
            996,
            1454,
            1002,
            1493
        ],
        "mask": {
            "size": [
                1860,
                2880
            ],
            "counts": "bVBiYjIzUGoxM04xTzAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDBPMk5UYmReMg=="
        },
        "automatic_annotation": false
    },

• After

{
        "token": "3dbe75ec855f9db4f35a1a457d88c7a0",
        "sample_data_token": "02382f82a702f04cdafa69d9eac0deed",
        "instance_token": "6ad49cbec66e7221dc2a3b6495bc7f9e",
        "category_token": "ff345f4048d27554367c25f35488510e",
        "attribute_tokens": [],
        "bbox": [
            1542,
            1815,
            1542,
            1852
        ],
        "mask": {
            "size": [
                2880,
                1860
            ],
            "counts": "VlZibzQxb2kyMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMGpVYjA="
        },
        "automatic_annotation": false
    },

test command

Reference

Notes for reviewer

[ktro2828]

Script to overwrite object_ann.json and surface_ann.json after transposing mask shape:

"""This file performs to convert segmentation mask shape from (height, width) order to (width, height) order.

USAGE:
    python3 convert_mask.py <DATA_ROOT>

NOTE:
`DATA_ROOT` is expected as follows:

```shell
DATA_ROOT/
├── <T4 Dataset2>
│   ├── annotations
│   ├── data
│   ...
├── <T4 Dataset2>
│   ├── annotations
│   ├── data
│   ...

"""

from future import annotations

import argparse
import base64
from glob import glob
import json
import os.path as osp
from typing import Any

from nuimages import NuImages
import numpy as np
import pycocotools.mask as cocomask

def convert(filepath: str) -> None:
"""Convert mask shape from (height, width) order to (width, height) order in a single dataset.

After convering mask shape, the original `object_ann.json` and `surface_ann.json` will be overwritten by modified files.

Args:
    filepath (str): Root path to the single T4 dataset.
"""
nuim = NuImages("annotation", filepath)

new_object_ann_list: list[dict[str, Any]] = []
for object_ann in nuim.object_ann:
    new_object_ann = object_ann.copy()

    mask = cocomask.decode(new_object_ann["mask"])
    rle = cocomask.encode(np.asfortranarray(mask.T))
    rle["counts"] = base64.b64encode(rle["counts"]).decode("ascii")
    new_object_ann["mask"] = rle

    new_object_ann_list.append(new_object_ann)

new_surface_ann_list: list[dict[str, Any]] = []
for surface_ann in nuim.surface_ann:
    new_surface_ann = surface_ann.copy()

    mask = cocomask.decode(new_surface_ann["mask"])
    rle = cocomask.encode(np.asfortranarray(mask.T))
    rle["counts"] = base64.b64encode(rle["counts"]).decode("ascii")
    new_surface_ann["mask"] = rle

    new_surface_ann_list.append(new_surface_ann)

with open(osp.join(filepath, "annotation/object_ann.json"), "w") as f:
    json.dump(new_object_ann_list, f, indent=4)

with open(osp.join(filepath, "annotation/surface_ann.json"), "w") as f:
    json.dump(new_surface_ann_list, f, indent=4)

def main() -> None:
"""Run process."""
parser = argparse.ArgumentParser()
parser.add_argument(
"data_root",
type=str,
help="Root direcotpy path storing target T4 datasets.",
)
args = parser.parse_args()

filepaths = glob(osp.join(args.data_root, "*"))
for filepath in filepaths:
    convert(filepath)

if name == "main":
main()