Traffic Object Detection for BFMC

This repository contains scripts to train an object detection model for the Bosch Future Mobility Challenge (BFMC): https://boschfuturemobility.com/

The model can detect and classify the following:

• Traffic signs: oneway, stop, highway entrance/exit, roundabout, parking, crosswalk, no entry, priority
• Cars
• Pedestrians

It includes tools for annotation, label generation using an existing model, preprocessing, synthetic dataset generation, training, evaluation, and analysis.

---

🛠 Installation

Install dependencies:

pip install -r requirements.txt

---

Pipeline

1. Clone the repository

   cd ~
   git clone https://github.com/simonli357/TrafficObjectDetection.git

2. Download the dataset

   Download the bfmc_data.zip from this Google Drive link: click here

3. Unzip and create a symbolic link to the dataset

   Linux:

   ln -s /path/to/unzipped/dataset ~/TrafficObjectDetection/bfmc_data

   Windows (PowerShell):

   New-Item -ItemType SymbolicLink -Path "C:\Users\YourUser\TrafficObjectDetection\bfmc_data" -Target "D:\Path\To\Unzipped\Dataset"

4. Generate datasets A and C:

   a) Generate augmented cropped images:

   python3 preprocessing/apply_augmentations.py

   (Tune values in the script to your needs)

   b) Resize the images with Gaussian distribution:

   python3 preprocessing/resize_normal.py

   c) Create dataset A:

   python3 preprocessing/create_datasets_a.py

   d) Create dataset C:

   python3 preprocessing/create_datasets_c.py

5. Combine all datasets into one

   python3 preprocessing/combine_datasets.py

6. Train the model

   python3 training/train.py

7. Test all epochs and save results

   python3 training/test_all_epochs.py

8. Plot results for all epochs

   python3 training/plot_all_epochs.py

---

Annotation Tools

1. Generate labels for annotation

   python3 annotation/generate_labels.py

   (Modify the script to point to the images you want to annotate)

2. Adjust labels manually

   python3 annotation/correct_labels.py

   This will open a GUI window with the image and bounding boxes:

   • Move or resize bounding boxes by clicking on them
   • Add new box: press a and draw
   • Change label class:
     • 0 to 9 → respective classes
     • i → class 10
     • o → class 11
     • p → class 12
   • Save: s
   • Navigate: b (back), n (next)
   • Crop and save bounding boxes: c
   • Hide part of the image: h and draw a box
   • Delete image and label: x
   • Quit: q