Model Training Repository

This repository is designed for training YOLO5, YOLO7 object detection models.

Model Results

YOLOv5, YOLOv7, and Fine-tuned Model Results

Prerequisites

1. Pull Docker Image

Before training, please pull the Docker image which contains pre-trained YOLO5 and YOLO7 models:

docker pull timttu/yolov5-nrp:v6

Note: The Docker image already includes pre-trained YOLO5 and YOLO7 models. If you don't want to train from scratch, you can directly use these pre-trained models.

2. Run Preprocessing Script

Before running any training scripts, you must first run the preprocessing script to download and prepare the datasets:

bash preprocessing.sh

This script will automatically download and extract:

• Training Dataset: car_train_split

  • Google Drive: https://drive.google.com/uc?id=1qL-1PV1jvNDRF_yToKUGptVpvmZUfor1

• Test Dataset: car_test

  • Google Drive: https://drive.google.com/uc?id=1E5mqA18Dto2l0MjqERjQm3BSAPbcwoeF

• Fine-tune Dataset: finetune-all

  • Google Drive: https://drive.google.com/uc?id=1pbHF8CLRT3vCcjA9fyFiDEqciqvOn0qM

The preprocessing.sh script will skip downloading if the datasets already exist on your machine.

Using Pre-trained Models

After building the Docker image with the models included, you can use the pre-trained models for testing:

Model Locations in Docker Container:

• YOLOv5 models: /workspace/yolov5/yolo5_best.pt and /workspace/yolov5/yolo5_fintune_best.pt
• YOLOv7 model: /workspace/yolov7/yolo7_best.pt

To run testing:

1. Update the dataset path in test/test_253_dataset.yaml to point to your mounted dataset location

2. Run the test script:

   docker run --rm --gpus all \
     -v /path/to/your/datasets:/workspace/datasets \
     -v $(pwd)/test:/workspace/test \
     timttu/yolov5-nrp:v6 \
     python /workspace/test/test_253_local.py

3. Modify test/test_253_dataset.yaml as needed:

   • Change the path: field to match your dataset location
   • Adjust train: and val: paths if needed

Project Structure

• run_yolo5s.py - YOLO5 training script
• run_yolo7.py - YOLO7 training script
• test_yolo5s.py - YOLO5 testing script
• preprocessing.sh - Dataset preparation script
• Car_model_*.yaml - Model configuration files
• split_dataset/ - Dataset splitting utilities
• job & pod/ - Kubernetes configuration files
• YOLO5s_result/ - YOLO5 training results
• YOLO7_result/ - YOLO7 training results