This project demonstrates how to use Machine Learning Environment to make a machine–learning model that identifies hand gestures and makes the Mars Rover move accordingly.
We are going to use the Hand Classifier of the Machine Learning Environment. The model works by analyzing your hand position with the help of 21 data points.
Hand Gesture Classifier Workflow
Follow the steps below:
- Open PictoBlox and create a new file.
- Select the coding environment as appropriate Coding Environment.
- Select the “Open ML Environment” option under the “Files” tab to access the ML Environment.
- Click on “Create New Project“.
- A window will open. Type in a project name of your choice and select the “Hand Gesture Classifier” extension. Click the “Create Project” button to open the Hand Pose Classifier window.
- You shall see the Classifier workflow with two classes already made for you. Your environment is all set. Now it’s time to upload the data.
Class in Hand Gesture Classifier
There are 2 things that you have to provide in a class:
- Class Name: It’s the name to which the class will be referred as.
- Hand Pose Data: This data can either be taken from the webcam or by uploading from local storage.
Note: You can add more classes to the projects using the Add Class button.
Adding Data to Class
You can perform the following operations to manipulate the data into a class.
- Naming the Class: You can rename the class by clicking on the edit button.
- Adding Data to the Class: You can add the data using the Webcam or by Uploading the files from the local folder.
- Webcam:
Note: You must add at least 20 samples to each of your classes for your model to train. More samples will lead to better results.
Training the Model
After data is added, it’s fit to be used in model training. In order to do this, we have to train the model. By training the model, we extract meaningful information from the hand pose, and that in turn updates the weights. Once these weights are saved, we can use our model to make predictions on data previously unseen.
The accuracy of the model should increase over time. The x-axis of the graph shows the epochs, and the y-axis represents the accuracy at the corresponding epoch. Remember, the higher the reading in the accuracy graph, the better the model. The range of the accuracy is 0 to 1.
Testing the Model
To test the model, simply enter the input values in the “Testing” panel and click on the “Predict” button.
The model will return the probability of the input belonging to the classes.
Export in Python Coding
Click on the “Export Model” button on the top right of the Testing box, and PictoBlox will load your model into the Python Coding Environment if you have opened the ML Environment in Python Coding.
Code
The following code appears in the Python Editor of the selected sprite.
####################imports####################
# Do not change
import numpy as np
import tensorflow as tf
import time
sprite=Sprite('Tobi')
import time
quarky = Quarky()
rover = MarsRover(4, 1, 7, 2, 6)
# Do not change
####################imports####################
#Following are the model and video capture configurations
# Do not change
model=tf.keras.models.load_model(
"num_model.h5",
custom_objects=None,
compile=True,
options=None)
pose = Posenet() # Initializing Posenet
pose.enablebox() # Enabling video capture box
pose.video("on",0) # Taking video input
class_list=['forward','backward','left','right','stop'] # List of all the classes
def runQuarky(predicted_class):
if pose.ishanddetected():
if predicted_class == "forward":
rover.home()
rover.setinangle(0)
quarky.runtimedrobot("F",100,3)
if predicted_class == "backward":
rover.home()
rover.setinangle(0)
quarky.runtimedrobot("B",100,3)
if predicted_class == "left":
rover.home()
rover.setinangle(40)
quarky.runtimedrobot("L",100,3)
if predicted_class == "right":
rover.home()
rover.setinangle(40)
quarky.runtimedrobot("R",100,3)
if predicted_class == "stop":
quarky.stoprobot()
else:
quarky.stoprobot()
# Do not change
###############################################
#This is the while loop block, computations happen here
# Do not change
while True:
pose.analysehand() # Using Posenet to analyse hand pose
coordinate_xy=[]
# for loop to iterate through 21 points of recognition
for i in range(21):
if(pose.gethandposition(1,i,0)!="NULL" or pose.gethandposition(2,i,0)!="NULL"):
coordinate_xy.append(int(240+float(pose.gethandposition(1,i,0))))
coordinate_xy.append(int(180-float(pose.gethandposition(2,i,0))))
else:
coordinate_xy.append(0)
coordinate_xy.append(0)
coordinate_xy_tensor = tf.expand_dims(coordinate_xy, 0) # Expanding the dimension of the coordinate list
predict=model.predict(coordinate_xy_tensor) # Making an initial prediction using the model
predict_index=np.argmax(predict[0], axis=0) # Generating index out of the prediction
predicted_class=class_list[predict_index] # Tallying the index with class list
print(predicted_class)
runQuarky(predicted_class)
# Do not change
Logic
- If the identified class from the analyzed image is “forward,” the Mars Rover will move forward at a specific speed.
- If the identified class is “backward,” the Mars Rover will move backward.
- If the identified class is “left,” the Mars Rover will move left.
- If the identified class is “right,” the Mars Rover will move right.
- Otherwise, the Mars Rover will be in the home position , stopped.
def runQuarky(predicted_class):
if pose.ishanddetected():
if predicted_class == "forward":
rover.home()
rover.setinangle(0)
quarky.runtimedrobot("F",100,3)
if predicted_class == "backward":
rover.home()
rover.setinangle(0)
quarky.runtimedrobot("B",100,3)
if predicted_class == "left":
rover.home()
rover.setinangle(40)
quarky.runtimedrobot("L",100,3)
if predicted_class == "right":
rover.home()
rover.setinangle(40)
quarky.runtimedrobot("R",100,3)
if predicted_class == "stop":
quarky.stoprobot()
else:
quarky.stoprobot()
Output
