Crowd counting’s documentation!

Crowd_counting is a library to train and use your own crowd counting models using python3 and based on tensorflow. It was developped in the context of the master Thesis of Louis Robins and Henri Collin at UCLouvain. The source code is available on github .

The aim of this documentation is to guide anyone that would like to

• Train a crowd counting model based on the CSRNet architecture on its own datas

• Use a pretrain model to predict a density map and the number of people on an image

The CSRNet model implementation is highly inspired from the implementation of the creators . Some changes were made to adapt it to Python 3.

Guides:

• Installation
• Train a new model
  • Folder structure
  • Data augmentation
  • Ground_truth generation
  • Model training
  • Model evaluation
• Load model and make predictions
  • Available pretrained Models
  • Loading a model
  • Make predictions

Modules:

• Data augmentation Module
• Ground_truth generation Module
• Training Module
• Model Module

Indices and tables

• Index

• Module Index

• Search Page