Street Level Imagery Labeler (SLIL)

What is it?

The SLIL tool was created to help in the task of manually labelling street level images into the categories of images with:

Trees
Overhead power lines
Intersections of power lines with trees

How to install it?

It is based on python 3.6, but theoretically it should work on any newer version.

So firstly you need to install the packages from the requirements.txt by running for example:

pip3 install -r requirements.txt

Then you can run the code running:

python3 slil.py

Ok, it runs, now what?

Well, you can load a folder using the menu:

File -> Choose folder

then choose a folder containing images.

Notice that the summary file (by default summan.smr) will be created (or read from) this folder. You can also load a custom file by pressing:

File -> Load labels

Then you can check the checkboxes corresponding to classes present in the current image (by defaul trees, power lines or the intersection of them) and either press the yellow Save button, of if the checkbox Auto-save is checked, you can also just press the yellow Next -> button or its corresponding keyboard key right arrow ->.

After labeling some images you need to commit the chages back to the summary file by using the Save button:

1. File -> Save labels

Or alternatively you can create another summary file elsewhere using:

1. File -> Save as labels

And the other checkboxes and buttons?

The yellow <- Previous button selects the previous indexed image or the last in the list if the current one is the first (image with the index 0).
The Only unlabeled checkbox skips images that were already labeled.

Warning: If you keep 'Auto-save' and 'Only unlabeled' both checked, and accidentally press the 'Next' or 'Previous' button will result in the current labels being saved and an attempt to come back to fix it will skip it because now that image is no longer unlabelled. If you need to come back to an already labeled image firstly you need to uncked the 'Only unlabeled' checkbox.

The Ignore exported checkbox will skip images marked as exported.
The menu Export has the buttons:
Export batch: Used to create another summary file with a part (i.e. batch size) of the unlabelled images. This is usefull to allow multiple persons to simultaneously label the images, provided that each person has access to the partition that the exported summary file refers to.
Merge (import batch): After annotating the exported summary file it can be merged back into the original summary using this button. This will take the annotated entries from the exported file and write their values over the corresponding entries in the current opened summary file. Notice that entries already fulfilled in the current summary file will be ignored. Furthermore, entries that only exist in the exported file will also be ignored.

Ok, but can I annotate other things?

Yes, you can! But be careful and avoid opening a summary file that has different labels.

To change the labels showing in the application you can set the variable labels in the file summarymanager_config.txt. That is a comma-separated list, so you can change the current labels, or insert new ones by appending a new label with a comma. For example, you can add the label 'Car' after 'Intersetion' by setting labels from

labels = 'Tree,Pole w/ wire,Intersection'

labels = 'Tree,Pole w/ wire,Intersection,Car'.

Citation

If you find the code helpful in your resarch or work, please cite the following software.

@software{Oliveira_Street_Level_Image_2020,
author = {Oliveira, Artur André and Hirata Jr., Roberto},
month = aug,
title = {{Street Level Image Labeler (SLIL)}},
url = {https://github.com/arturandre/SLIL},
version = {1.0.0},
year = {2020}
}