Detecting Traffic Flows

Getting started

1. This repo contains two submodules, yolov5 [1] and sahi [2]. After cloning this repo run git submodule update to download the and update these repos.

[1] https://github.com/ultralytics/yolov5

[2] https://github.com/obss/sahi

2. Make sure the symlinks pointing between folders are there. If not run

cd inference 
ln -s ../sahi/sahi sahi
ln -s ../yolov5/models models
ln -s ../yolov5/utils utils

3. Download yolov5 weights to inference/trained_models/yolov5/ and tood weights to inference/trained_models/tood/.

4. Environment setup:

   Inference:

   conda create --name inference
   conda activate inference
   conda install pytorch=1.10.0 torchvision=0.11.1 cudatoolkit=11.3 -c pytorch
   cd inference
   pip install -r trained_models/tood/tood-req.txt
   pip install -r ../sahi/requirements.txt
   pip install yolov5

   tf:

   11.2.1-cudnn8.1.0.77
   source /vol/cuda/11.2.1-cudnn8.1.0.77/setup.sh
   
   

Inference

To detect traffic on images make sure to have the images inside inference/WV3 and then run cd inference and python inference_yolov5.py or python inference_tood.py for the (worse peforming) tood detector.

To detect traffic run without image exports

python inference.py --conf_thresh 0.05 --novisual --export_pickle --name single-lane-yolov5
python inference.py --conf_thresh 0.05 --novisual --export_pickle --source_image_dir WV3-dual-lane --name double-lane-yolov5

Traffic Count Prediction

Once inference on the images has been peformed Traffic Count predictions can be made. Please see the notebook Traffic-Count-Pipeline in the folder traffic-flow for a step by step walkthrough of the inference process. Before running the notebook Traffic Count data needs downloading. The files download_high_quality_reports.py, download_image_time_reports.py, download_image_year_reports.py must be run.

AADT predictions

The linear model needs around 100GB of RAM to run, the dataset is huge! To produce the model run Flow-Model-Baseline.py. Afterwards the notebooks Flow-Model and Flow-Model-Linear walkthrough the prediction of AADT/AMD with median and linear models respectively.

Plots

This is a miscellanious notebook used to create plots for my report. I have left it in for the curious to explore :) .