vaibhav-vemula/Pedestrian_Detection_DVCPipeline

DVC pipeline for Object Detection using YoloV5 with Data Versioning and Transforming

Pedestrian Detection & Segmentation MLOps Pipeline

DVC pipeline for Pedestrian Detection & Segmentation using YoloV5 with Data Versioning

Checkout to clouds3 branch for pushing results/outputs to AWS S3 Bucket

Getting Started

1. Create a Python environment

python3 -m venv <env_name>
source <env_name>/bin/activate

2. To initialize DVC and GIT

pip3 install dvc
git init
dvc init

3. Installing dependencies

To install requirements for running object detection pipeline with YoloV5 Requires PyTorch, CUDA(if GPU Enabled)

pip3 install -r requirements.txt

4. Setting paramenters

Here dcount is the number of versions of datasets uploaded(dcount = 0 when initialised)

# file params.yaml
ingest:
    dcount:0

5. To run DVC pipeline

dvc repro

6. To run Streamlit App

streamlit run main.py

7. Adding pipeline stage

dvc run -n <Stage_name> 
    -p ingest.dcount -p <add parameters> 
    -d src/<Stage_file>.py -d <Any dependencies>
    -o data/<Output dir> 
    python3 src/prepare.py

8. Pipeline DAG

dvc dag

9. Push results/outputs to AWS S3

Create .streamlit/secrets.toml file and add your AWS S3 BUCKET ACCESS_KEY and SECRET_KEY

ACCESS_KEY = 'xxxxxxx'
SECRET_KEY = 'xxxxxxxxxxxxxx'