Raspberry Pi deployment in Computer Vision - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to load a TensorFlow Lite model on Raspberry Pi.

Computer Vision

import tflite_runtime.interpreter as tflite

interpreter = tflite.Interpreter(model_path='model.tflite')
interpreter.[1]()

Drag options to blanks, or click blank then click option'

Acompile

Brun

Callocate_tensors

Dload_model

Attempts:

3 left

💡 Hint

Common Mistakes

Using load_model() which is not a method of the interpreter.

Trying to run inference before allocating tensors.

The method allocate_tensors() prepares the model's tensors for inference on Raspberry Pi.

2fill in blank

medium

Complete the code to set the input tensor for the model on Raspberry Pi.

Computer Vision

input_details = interpreter.get_input_details()
input_data = ...  # preprocessed input image
interpreter.[1](input_details[0]['index'], input_data)

Drag options to blanks, or click blank then click option'

Aset_tensor

Binput_tensor

Cset_input_tensor

Dfeed_tensor

Attempts:

3 left

💡 Hint

Common Mistakes

Using a non-existent method set_input_tensor().

Confusing input_tensor as a method instead of a key.

The method set_tensor() sets the input tensor data for the interpreter.

3fill in blank

hard

Fix the error in the code to run inference on Raspberry Pi.

Computer Vision

interpreter.[1]()
output_details = interpreter.get_output_details()
output_data = interpreter.get_tensor(output_details[0]['index'])

Drag options to blanks, or click blank then click option'

Arun_inference

Binvoke

Cexecute

Dpredict

Attempts:

3 left

💡 Hint

Common Mistakes

Using run_inference() which does not exist.

Using execute() or predict() which are not TensorFlow Lite methods.

The correct method to run inference is invoke() in TensorFlow Lite interpreter.

4fill in blank

hard

Fill both blanks to preprocess an image for Raspberry Pi model input.

Computer Vision

from PIL import Image
import numpy as np

image = Image.open('input.jpg').resize([1])
input_data = np.expand_dims(np.array(image), axis=[2]).astype(np.float32)

Drag options to blanks, or click blank then click option'

A(224, 224)

B0

C1

D(128, 128)

Attempts:

3 left

💡 Hint

Common Mistakes

Using wrong image size like (128, 128).

Expanding dims on axis 1 instead of 0.

The image is resized to (224, 224) and expanded along axis 0 to add batch dimension.

5fill in blank

hard

Fill all three blanks to postprocess the model output on Raspberry Pi.

Computer Vision

output_details = interpreter.get_output_details()
output_data = interpreter.get_tensor([1])
predicted_label = np.argmax(output_data[[2]])
confidence = output_data[[3]][predicted_label]

Drag options to blanks, or click blank then click option'

Aoutput_details[0]['index']

B0

D1

Attempts:

3 left

💡 Hint

Common Mistakes

Using wrong tensor index or batch index.

Mixing up batch indices 0 and 1.

We get the output tensor index, then use batch 0 to find predicted label and confidence.