csc4700-embedded-collision-detection

• Perform binary classification (Collision / No collision) using vehicle acceleration data (x,y,z axis acceleration) collected using accelerometer sensor within 3-seconds timeframe.
• Entries collected within same timeframe is identified by an unique batch_id
• 5 entries of data within 3-seconds timeframe form a complete journey. All entries in the same journey will always have same is_collision value.
• Display real-time data collected and prediction made on ThinkSpeak dashboard (Note: ThinkSpeak has limitation of 15 seconds between requests sent)

How to launch Trained-Model-Hosting Server

1. Activate virtual env (if any)

source venv/Scripts/activate

2. Install required packages (for first time user)

pip install -r requirements.txt

3. Run server

uvicorn server:app --host 0.0.0.0 --port 8000

Hardware & Software Used

• ADXL345 Accelerometer
• ESP-32 Wifi Module
• ThinkSpeak (Dashboard & store dataset collected)
• Google Colab (ML)

TODO

Code

• Collect accelerometer data and upload to ThinkSpeak
• Preprocess dataset downloaded from ThinkSpeak (2 channel, train data channel and test data channel)
• Train and evaluate classification model
• Host trained model using FastAPI
• Use trained model to make real-time prediction
• Display data and graphs on ThinkSpeak dashboard

Non-Code

• tinkercard architectural diagram
• flowchart for methodology
• architectural diagram
• report and presentation slide