road-speeds

Predicting road speeds over time using satellite imagery and tabular data

Installation Instructions

Step 1: Set Up Python Virtual Environment

Ensure you have Python 3.9 installed on your system. If not, install it from the official Python website.
You may need to install the python3-venv package to create a virtual environment, which can be done using:

sudo apt install python3-venv

Create a virtual environment named road_env using:

python3.9 -m venv road_env

Step 2: Activate the Virtual Environment

Activate the virtual environment with:

source road_env/bin/activate # On Unix or MacOS .\road_env\Scripts\activate # On Windows

Step 3: Install Required Dependencies

Before installing the python packages, some packages require the Python development package and GDAL requires the libgdal library to be pre-installed on your system. Install these using:

sudo apt-get update sudo apt-get install libgdal-dev sudo apt install python3.9-dev

With the virtual environment activated, install the dependencies (ensure 'requirements.txt' is in the current directory) by running:

pip install -r requirements.txt

Step 4: Verify Installation

You can verify the installation by checking the Python version and installed packages using:

python --version pip list

Environment Setup

Step 1: Enter the Repository

Make sure you are in the project repository

Step 2: Set Up the Python Environment

[Follow the installation instructions provided in installation_requirements.txt to set up the Conda environment.]
Ensure the environment road_env is activated before proceeding to the next steps (use: source road_env/bin/activate)

Step 3: Data

In order to generate the 224 x 224 crops of the road segments (which are used as input for the models), the generate_crops.py script needs to be run. generate_crops.py expects the following directory structure:

PROJECT_ROOT │
├── data
│ ├── GPStracks_forDLcourse.csv # GPS tracks CSV file
│ ├── img_id_map.json # Generated JSON file (output by script)
│ ├── crop_info_base.csv # Generated crops info CSV file (output by script)
│ │
│ ├── sat_tifs # Directory for TIFF non-sequence images
│ │ ├── [various TIFF files]
│ │
│ ├── sat_tifs_2 # Directory for TIFF sequence images
│ │ ├── batch2
│ │ ├── R1_[various TIFF files]
│ │ ├── R2_[various TIFF files]
│ │ └── R3_[various TIFF files]
│ │
│ ├── rgb_crops_mo_4 # Directory for RGB image crops (output by script)
│ │ ├── [generated RGB crop files]
│ │
│ └── nir_crops_mo_4 # Directory for NIR image crops (output by script)
│ ├── [generated NIR crop files]
│
├── [other project files and directories]
│
└── generate_crops.py

The GPStracks_forDLcourse.csv is found in the bucket associated with the project data and must be added to the data directory.

Generate the crops by running:

python generate_crops.py

Experiment Running

Step 0: Ensure proper environment

Before proceeding with running main.sh, ensure that the environment is set up as detailed in installation_requirements.txt and environment_setup.txt.

Step 1: Running the experiments

Running the experiments is done by running the following command in the terminal:

chmod +x main.sh ./main.sh

Here is the directory structure expected by seq_nn_train.py:

PROJECT_ROOT
│
├── data
│ ├── crop_info_full_4.csv # The main dataset CSV file
│ ├── [Other data files]
│
├── dataset
│ ├── init.py
│ ├── road_mean_std_dataset.py # Contains RoadMeanSTDDataset class
│ ├── road_mean_std_seq_dataset.py # Contains RoadMeanSTDSeqDataset class
│ ├── [Other dataset files]
│
├── models
│ ├── init.py
│ ├── resnet_concat_rnn.py # Contains the ResNetConcatRNN model
│ ├── [Other model files]
│
├── results
│ ├── models # Directory to save model outputs
│ │ ├── [Saved model files will be placed here]
│
├── transforms.py # Contains data transformations
│
├── util.py # Utility functions for the model
│
├── crop_info_norm_params.yaml # Normalization parameters
│
├── t_cols_norm.yaml # Names of normalized tabular feature columns
│
├── training_script.py # The main training script
│
└── [Other project files and directories]

vivekvajipey/road-speeds