Rellis3D-5Labels / 2_RELLIS-3D_split.py
VJ
Upload 4 files
d2d323b verified
raw
history blame
5.86 kB
# -*- coding: utf-8 -*-
"""
Created on Wed Sep 21 08:17:42 2022
----------------------------------------------------------------------------------
This script split the rellis 3D dataset into train val and test folders
----------------------------------------------------------------------------------
1. Create train, val and test folders in data/rellis folder
2. Create a new empty folder named as images and masks_id in train and val folders.
3. Create a new empty foldr named as images inside test folder.
4. Execute this script to copy images and masks_id to train folder.
5. Execute this script to copy images and masks_id to val folder.
6. Execute this script to copy images to test folder.
Output:
Coping images, please wait...
Total number of copied images: 3302
Images stored saved in: ./train//images
Coping masks_id, please wait...
Total number of copied masks: 3302
Images stored saved in: ./train/masks_id
Coping images, please wait...
Total number of copied images: 983
Images stored saved in: ./val//images
Coping masks_id, please wait...
Total number of copied masks: 983
Images stored saved in: ./val/masks_id
Coping images, please wait...
Total number of copied images: 1672
Images stored saved in: ./test//images
_-----------------------------------------------------------------------------
Mask_color images to instance segmentation
1. Create train, val and test folders in data/rellis folder
2. Create a new empty folder named as images and masks_color in train and val folders.
3. Create a new empty foldr named as images inside test folder.
4. Execute this script to copy masks_colors to train folder.
5. Execute this script to copy masks_colors to val folder.
6. Execute this script to copy masks_colors to test folder.
Output:
Coping images, please wait...
Total number of copied images: 3302
Images stored saved in: ./train//masks_colors
Coping images, please wait...
Total number of copied images: 983
Images stored saved in: ./val//masks_colors
'''
"""
import os
import pandas as pd
from shutil import copy
def search_image(count, imagename, path, DEST_PATH):
error=0
for filename in os.listdir(path):
if (filename.endswith(".bmp")) and (filename.startswith(imagename)):
copy(path +'/'+ filename, DEST_PATH+'/'+filename)
count+=1
error=0
#print(count)
break
else:
error=1
return count, error
DEST_PATH='train'
#DEST_PATH='val'
#DEST_PATH='test'
# Rellis 5 categories
pd_card_images = pd.read_csv('train5.csv',sep=";", names=['images','masks_id'], encoding ='latin1', low_memory=False)
#pd_card_images = pd.read_csv('val5.csv',sep=";", names=['images','masks_id'], encoding ='latin1', low_memory=False)
#pd_card_images = pd.read_csv('test5.csv',sep=";", names=['images','masks_id'], encoding ='latin1', low_memory=False)
# Rellis 6 categories
#pd_card_images = pd.read_csv('train.csv',sep=";", names=['images','masks_id'], encoding ='latin1', low_memory=False)
#pd_card_images = pd.read_csv('val.csv',sep=";", names=['images','masks_id'], encoding ='latin1', low_memory=False)
#pd_card_images = pd.read_csv('test.csv',sep=";", names=['images','masks_id'], encoding ='latin1', low_memory=False)
# Rellis 20 labels
#pd_card_images = pd.read_csv('train20.csv',sep=";", names=['images','masks_id'], encoding ='latin1', low_memory=False)
#pd_card_images = pd.read_csv('val20.csv',sep=";", names=['images','masks_id'], encoding ='latin1', low_memory=False)
#pd_card_images = pd.read_csv('test20.csv',sep=";", names=['images','masks_id'], encoding ='latin1', low_memory=False)
#pd_card_images = pd.read_csv('test_maskcolors.csv',sep=";", names=['maskcolors'], encoding ='latin1', low_memory=False)
# Rellis 20 masks_colors for instance segmentation
#pd_card_images = pd.read_csv('train_color20.csv',sep=";", names=['images', 'masks_colors'], encoding ='latin1', low_memory=False)
#¶pd_card_images = pd.read_csv('val_color20.csv',sep=";", names=['images', 'masks_colors'], encoding ='latin1', low_memory=False)
#pd_card_images = pd.read_csv('test_color20.csv',sep=";", names=['images', 'masks_colors'], encoding ='latin1', low_memory=False)
pd_card_images = pd_card_images.drop(labels=0, axis=0)
rpd_card_images=pd_card_images.reset_index(drop=True)
image_path=rpd_card_images['images']
mask_path=rpd_card_images['masks_id']
#maskcolors_path=rpd_card_images['masks_colors']
print('Coping images, please wait...')
count=0
for im in image_path:
filename = str(im).split("/")
txt_filename=str(filename[1])
#print(im)
copy('images/'+ im, DEST_PATH +'/images/'+txt_filename)
count+=1
print('Total number of copied images: ', count)
print('Images stored saved in: ', DEST_PATH+'/images')
# for idx, im in enumerate(maskcolors_path):
# filename = str(im).split("/")
# txt_filename=str(filename[1])
# orig_image = image_path[idx]
# copy('./maskscolors/' + orig_image, DEST_PATH +txt_filename)
# count+=1
# print('Total number of copied images: ', count)
# print('Images stored saved in: ', DEST_PATH+'/masks_colors')
## Only for train and val **************************************
print('Coping masks_id, please wait...')
count=0
for mask in mask_path:
filename = str(mask).split("/")
txt_filename=str(filename[1])
#print(mask)
copy('masks_id/'+ mask, DEST_PATH+'/masks_id/'+txt_filename)
count+=1
print('Total number of copied masks: ', count)
print('Images stored saved in: ', DEST_PATH+'masks_id')