LSD-seg label conversion

convert_label.py

import argparse
import json
import os

import numpy as np
from PIL import Image
from tqdm import tqdm

def swap_labels(np_original_gt_im, class_convert_mat):
    np_processed_gt_im = np.zeros(np_original_gt_im.shape)
    for swap in class_convert_mat:
        ind_swap = np.where(np_original_gt_im == swap[0])
        np_processed_gt_im[ind_swap] = swap[1]
    processed_gt_im = Image.fromarray(np.uint8(np_processed_gt_im))
    return processed_gt_im


def convert_citylabelTo16label():
    with open('data/synthia2cityscapes_info.json', 'r') as f:
        paramdic = json.load(f)

    class_ind = paramdic['city2common']

    city_gt_dir = "data/cityscapes/gtFine"
    split_list = ["train", "test", "val"]

    original_suffix = "labelIds"
    processed_suffix = "label16IDs"

    for split in tqdm(split_list):
        base_dir = os.path.join(city_gt_dir, split)
        place_list = os.listdir(base_dir)
        for place in tqdm(place_list):
            target_dir = os.path.join(base_dir, place)
            pngfn_list = os.listdir(target_dir)
            original_pngfn_list = [x for x in pngfn_list if original_suffix in x]

            for pngfn in tqdm(original_pngfn_list):
                gt_fn = os.path.join(target_dir, pngfn)
                original_gt_im = Image.open(gt_fn)

                processed_gt_im = swap_labels(np.array(original_gt_im), class_ind)
                outfn = gt_fn.replace(original_suffix, processed_suffix)
                processed_gt_im.save(outfn, 'PNG')


def convert_synthialabelTo16label():
    with open('data/synthia2cityscapes_info.json', 'r') as f:
        paramdic = json.load(f)

    class_ind = np.array(paramdic['synthia2common'])

    synthia_gt_dir = "data/RAND_CITYSCAPES/GT"

    # original_dir = os.path.join(synthia_gt_dir, "LABELS") # Original dir but this contains strange files
    original_dir = "data/segmentation_annotation/SYNTHIA/GT/parsed_LABELS"  # Not original. Downloaded from http://crcv.ucf.edu/data/adaptationseg/ICCV_dataset.zip
    processed_dir = os.path.join(synthia_gt_dir, "LABELS16")

    original_pngfn_list = os.listdir(original_dir)

    for pngfn in tqdm(original_pngfn_list):
        gt_fn = os.path.join(original_dir, pngfn)
        original_gt_im = Image.open(gt_fn)
        processed_gt_im = swap_labels(np.array(original_gt_im), class_ind)
        outfn = os.path.join(processed_dir, pngfn)
        processed_gt_im.save(outfn, 'PNG')


if __name__ == '__main__':
    parser = argparse.ArgumentParser(description='Convert Label Ids')
    parser.add_argument('dataset', type=str, choices=["city", "synthia"])
    args = parser.parse_args()
    if args.dataset == "city":
        convert_citylabelTo16label()
    else:
        convert_synthialabelTo16label()

synthia2cityscapes_info.json

{
  "classes":16,
  "city2common":[
    [0, 255],
    [1, 255],
    [2, 255],
    [3, 255],
    [4, 255],
    [5, 255],
    [6, 255],
    [7, 0],
    [8, 1],
    [9, 255],
    [10, 255],
    [11, 2],
    [12, 3],
    [13, 4],
    [14, 255],
    [15, 255],
    [16, 255],
    [17, 5],
    [18, 255],
    [19, 6],
    [20, 7],
    [21, 8],
    [22, 255],
    [23, 9],
    [24, 10],
    [25, 11],
    [26, 12],
    [27, 255],
    [28, 13],
    [29, 255],
    [30, 255],
    [31, 255],
    [32, 14],
    [33, 15],
    [-1, 255]
  ],
  "synthia2common":[
    [0, 255],
    [1, 9],
    [2, 2],
    [3, 0],
    [4, 1],
    [5, 4],
    [6, 8],
    [7, 5],
    [8, 12],
    [9, 7],
    [10, 10],
    [11, 15],
    [12, 14],
    [13, 255],
    [14, 255],
    [15, 6],
    [16, 255],
    [17, 11],
    [18, 255],
    [19, 13],
    [20, 255],
    [21, 3],
    [22, 255]
  ],
  "city_label": [
    "unlabeled",
    "ego vehicle",
    "rectification border",
    "out of roi",
    "static",
    "dynamic",
    "ground",
    "road",
    "sidewalk",
    "parking",
    "rail track",
    "building",
    "wall",
    "fence",
    "guard rail",
    "bridge",
    "tunnel",
    "pole",
    "polegroup",
    "traffic light",
    "traffic sign",
    "vegetation",
    "terrain",
    "sky",
    "person",
    "rider",
    "car",
    "truck",
    "bus",
    "caravan",
    "trailer",
    "train",
    "motorcycle",
    "bicycle",
    "license plate"
  ],
  "synthia_label": [
    "void",
    "sky",
    "Building",
    "Road",
    "Sidewalk",
    "Fence",
    "Vegetation",
    "Pole",
    "Car",
    "Traffic sign",
    "Pedestrian",
    "Bicycle",
    "Motorcycle",
    "Parking-slot",
    "Road-work",
    "Traffic light",
    "Terrain",
    "Rider",
    "Truck",
    "Bus",
    "Train",
    "Wall",
    "Lanemarking"
  ],
  "common_label":[
    "road",
    "sidewalk",
    "building",
    "wall",
    "fence",
    "pole",
    "light",
    "sign",
    "vegetation",
    "sky",
    "person",
    "rider",
    "car",
    "bus",
    "motocycle",
    "bicycle"
  ],
  "palette":[
    [128,64,128],
    [244,35,232],
    [70,70,70],
    [102,102,156],
    [190,153,153],
    [153,153,153],
    [250,170,30],
    [220,220,0],
    [107,142,35],
    [70,130,180],
    [220,20,60],
    [255,0,0],
    [0,0,142],
    [0,60,100],
    [0,0,230],
    [119,11,32],
    [0,0,0]],
  "mean":[
    73.158359210711552,
    82.908917542625858,
    72.392398761941593],
  "std":[
    47.675755341814678,
    48.494214368814916,
    47.736546325441594]
}