Collision Avoidance
In this example we'll collect an image classification dataset that will be used to help keep
JetBot safe! We'll teach JetBot to detect two scenarios free and blocked. We'll use this AI classifier to prevent JetBot from entering dangerous territory.
Step 1 - Collect data on JetBot
- Connect to your robot by navigating to
http://<jetbot_ip_address>:8888 - Sign in with the default password
jetbot - Shutdown all other running notebooks by selecting
Kernel->Shutdown All Kernels... - Navigate to
~/Notebooks/collision_avoidance/ - Open and follow the
data_collection.ipynbnotebook
Tip
We provide a pre-trained model so you can skip to step 3 if desired. This model was trained on a limited dataset using the Raspberry Pi V2 Camera with wide angle attachment.
Step 2 - Train neural network
Option 1 - Train on Jetson Nano
- Connect to your robot by navigating to
http://<jetbot_ip_address>:8888 - Sign in with the default password
jetbot - In the Jupyter Lab tab, navigate to
~/collision_avoidance - Open and follow the
train_model_resnet18.ipynbnotebook
Option 2 - Train on other GPU machine
- Connect to a GPU machine with PyTorch installed and a Jupyter Lab server running
- Upload the collision avoidance training notebook to this machine
- Open and follow the
train_model_resnet18.ipynbnotebook
Step 3 - Optimize the model on Jetson Nano
- Connect to your robot by navigating to
https://<jetbot_ip_address>:8888 - Sign in with the default password jetbot
- Shutdown all other running notebooks by selecting Kernel -> Shutdown All Kernels...
- Navigate to
~/Notebooks/road_following - Open and follow the
live_demo_resnet18_build_trt.ipynbnotebook to optimize the model with TensorRT
Step 4 - Run live demo on JetBot
- Connect to your robot by navigating to
http://<jetbot_ip_address>:8888 - Sign in with the default password
jetbot - Shutdown all other running notebooks by selecting
Kernel->Shutdown All Kernels... - Navigate to
~/Notebooks/collision_avoidance - Open and follow the
live_demo_resnet18_trt.ipynbnotebook to run the optimized model
Caution
JetBot will physically move in this notebook, make sure it has enough space to move around.


