Crowd Counting Labeler

Fork from https://github.com/Elin24/cclabeler

Crowd Counting Labeler is a tool for labeling pedestrians in a images. It is a web application which could be deployed quickly, and provides two types label: box and point.

At present, it does not support multi-class labeling, which would be a future function.

How to run demo

1. clone the repository to local workspace:

git clone https://github.com/fabricejourdan/cclabeler.git

2. create Python environment with Django and Pillow:

conda create --name cclabeler python=3.9.7 django=4.0 pillow=8.4.0 pandas==1.3.5

3. cd to the directory, run django server:

python manage.py runserver 0.0.0.0:8000

4. login to the address in browser and enjoy it.

http://localhost:8000/

deploy new image data for labeling

Image data should be put in data/ fold

├─images (contain images to label)
│      1.jpg
│      2.jpg
│      3.jpg
│      .....
│
├─jsons (for each image, json file containing selected points and boxes)
│      1.json
│      2.json
│      3.json
|      .....
│
└─marks
        1.json
        2.json
        3.json
        .....

Set of examples for Demo

1. Three users

user1 , user2 and golden

Password are define in the file /users/{username}.json

2. Each user have 3 images

user1 : images from ShanghaiTech_B

user2 : images from ShanghaiTech_B

golden : images from CityUHK-X-BEV dataset (https://github.com/daizhirui/CityUHK-X-BEV)

For each user, the list of images to label is define in the file /users/{username}.json

Summary (statistics)

loin http://localhost:8000/summary for summary.