# STAIR IMAGE LABELING

Those who are volunteering update in this [Sheet](https://docs.google.com/spreadsheets/d/1iy2miaabhztJy09Fet1oolqwJKyHrQCdLt3fsaIT6VY/edit?usp=sharing) your name, the set from where you ll be labeling, and the image numbers you are labeling
We are working on developing an algorithm that can be used to make bots climb stairs. We need dataset to train.
To know more about our work visit our [Github Repo](https://github.com/IvLabs/Stair-Climber). Do star the repository!!

## Why is data genration and labeling important?
* Data Generation is an important task for researchers working in the fields of Computer Vision and Machine Learning.
* In order to do a new task there isn't dataset readily available so we need to take efforts to generate and label our own data set. Thus everone working in this field must know how to label images.

## Why is it benificial for you?
* You will gain an experience of labeling data.
* This data will be available for you and coming batches of IvLabs for training and testing.

## What you have to do?
We use label me for labeling. It is available [online](http://labelme.csail.mit.edu/Release3.0/) too but it is recommended to download and use.
### Installations

#### For windows
If you have anaconda run these commands in anaconda prompt.

# python3
conda create --name=labelme python=3.6
conda activate labelme
pip install labelme

You can install anaconda if you don't have it by searching up the [net](https://docs.conda.io/projects/conda/en/latest/user-guide/install/windows.html)

#### For UBUNTU

Refer [Github page of Labelme](https://github.com/wkentaro/labelme/tree/master/examples/tutorial)

### How to label
* First download the raw images in a folder. Remember the directory in which you have downloaded.
* Images
* [ Test Images Download From here](https://drive.google.com/drive/folders/1C_hs_SNoOylNDgFNjWD9RqRpPO9PH1Mn?usp=sharing)
* [Train Images Download from here](https://drive.google.com/drive/folders/131zvfr2t4kJQSvQFLU1Iay0CTmTr8x-r?usp=sharing)
* [Validation Images Download from here](https://drive.google.com/drive/folders/1c1ym5iVM02g4KKwtSaGkFvM-hBJ-nP0X?usp=sharing)
* Open anaconda prompt, Activate the env `conda activate labelme`
* Start the GUI run `labelme`
* Click on open directory, select the directory in which you have downloaded the images
* Follow the video to see how to label.
* Give label 1 to when prompted.
* You will get a json file on doing this
* Click next and label all the images.
> NOTE : Saved the .json file in directory same as that of raw images.


* After all this is done you will have to run the following command `labelme_json_to_dataset <image number>.json -o <image number>`
Ex I will run `labelme_json_to_dataset 2.json -o 2` because my file name is 2.json
> Note : Change the directory to the one in which you are storing json file, which should be same as that of the one in which raw images are kept if you have followed!!
* After doing this you will get a folder with name same as that of the json file. It will have 4 files in it. There will be a label.png file which will be a mask of the selection you have made in previous step.


* After you have generated all the folders you can delete anything **OUTSIDE THE GENERATED FOLDERS**
### Uploading
Upload the folders with names as that of the raw images that have were created in the above steps to the drive link below.
> Upload the images in the respective folders from where you have taken.

[Upload here](https://drive.google.com/drive/folders/1lH2vXkfFxdYQFUP9Nu-p1yCJWMtNeZS0?usp=sharing)

### Doubts
For doubts tag `@sushant kumar` or `@Aadithya Iyer` on general channel and ask