Creating a self-driving car by connecting a JETSON Nano board with a child-friendly car. I am designing and modifying artificial intelligence algorithms to implement autonomous driving for the car. I am also incorporating image recognition using YOLO, to stop the car on stop sign.
◇ HARDWARE SETUP
- JETBOT setup
- JETBOT & Car Motor setup
◇ SOFTWARE SETUP
- SD card & connect using wifi
- ROAD FOLLOWING
- road_following.zip
- rf_data_collection.ipynb
- rf_train_model.ipynb
- rf_live_detection.ipynb
- OBJECT DETECTION
- object_detection.zip
We transfered resnet18 to use it as a regression model. Training data was image with x,y coordinates as a label. x,y coordinate indicates where to stir the vehicle. We made 3000 images as a dataset. 10% of the dataset as test dataset and 90% as training dataset.
We use Yolo v5 to detect stop sign to stop the vehicle.
◇ Attempt1: data distribution
| total data | 1800 |
|---|---|
| straight | 800 |
| left | 600 |
| right | 400 |
◇ data sample
◇ Attempt2: data distribution
| total data | 1800 + 1200 |
|---|---|
| straight | 1000장 |
| left | 1000장 |
| right | 1000장 |
◇ data sample
◇ Attempt3: data distribution
| total data | 1800 + 1200 + 1500 |
|---|---|
| straight | 1000 + 1500 |
| left | 1000 + 0 |
| right | 1000 + 0 |
◇ data sample