alialamiidrissi / Trajnet

Trajnet tools and preprocessed datasets

This repository contains functions to preprocess and plot the datasets for the Trajnet challenge

How to generate the preprocessed data and the 'dataset_specs.csv' file ?

• To generate the preprocessed data, run the 'transformed_trajectories_generation.py' script as follow :

python transformed_trajectories_generation.py --output_folder [output_folder_path] --path_to_data [path_to_the_original_Trajnet_datasets]

• To generate the'dataset_specs.csv' file, run the 'csv_specs_generator.py' script as follow :

python csv_specs_generator.py --output_file [output_csv_file_name] --path_to_data [path_to_the_original_Trajnet_datasets]

Files description

• plotting_tools.py: Python functions to plots the datasets
• trajectories_trans_tools.py: Python functions to preprocess the datasets
• transformed_trajectories_generation.py: Python script to generate the preprocessed datasets
• csv_specs_generator.py: Python script to generate the 'stat_dataset' csv file
• Preprocessed_data.zip: Zip file containing the preprocessed datasets and has the following structure:
  • Every folder contains all trajectories for a given dataset
  • Every preprocessed trajectory is saved in a separate file named '[file]_[ped_id].txt'.
  • The first 20 frames of every file correspond to the preprocessed pedestrian trajectory with id = ped_id
  • The remaining entries correspond to the preprocessed pedestrian neighbors
• dataset_specs.csv: CSV file containing information about every pedestrian in every dataset.Its columns are described below:
  • Dataset: This column specifies the pedestrian dataset
  • File: This column specifies the pedestrian file
  • Track_ID: This column specifies the pedestrian ID
  • Mean_rotation: This column specifies the mean rotation of the pedesterian trajectory with respect to the vector between its first and second position
  • nb_neighbors : This column specifies the current pedestrian number of neighbors
  • nb_critical_neighbors : This column specifies the number of neighbors that are at most 6 meters far from the current pedestrian
  • mean_nb_neighbors_per_frame: This column specifies the mean number of neighbors of the current pedestrian per frame
  • mean_nb_critical_neighbors_per_frame: This column specifies the mean number of neighbors per frame that are at most 6 meters far from the current pedestrian
  • mean_contact : This columns specifies the mean number of frames during which the current pedestrians appears with a neighbor
  • is_static : This column specifies if the current pedestrian does not move