IPCV-QB [Unit–5,6]

1) Define image matching.

Ans - Image matching is the process of comparing
two or more images to determine if they are similar
or identical. It involves analysing the visual content of
images to find similarities based on features such as
colours, shapes, textures, and patterns. It's like
finding a needle in a haystack, but instead of needles
and hay, we're dealing with pixels and images.

2) Explain image matching after Erosion and Dilation

with example.
Ans - After erosion and dilation operations on images,
image matching involves comparing the resulting
images to identify similarities or differences.
**Erosion** shrinks the boundaries of objects in an
image by removing pixels based on the presence of
neighbouring pixels.
**Dilation** expands the boundaries of objects in an
image by adding pixels based on the presence of
neighbouring pixels.
For example, let's say we have an image of a circle.
After applying erosion, the circle becomes smaller,
with its edges smoothed out. After dilation, the circle
becomes larger, with its edges becoming more
defined.
In image matching, we compare these processed
images to original ones or other processed images to
find similarities or differences in shapes, textures, or
patterns, helping in tasks like object recognition or
image registration.
2) Explain image matching of human face with example.

Ans - Image matching of human faces involves

comparing facial features in images to determine if
they belong to the same person.
For example, let's consider two images of a person's
face. To match these faces, the system might analyse
features like the distance between the eyes, the
shape of the nose, and the contour of the lips.
If the analysed features are sufficiently similar, the
system can conclude that the faces belong to the
same person. This process is crucial in applications
like facial recognition for security systems or
identifying individuals in photographs.

4) Define object detection.

Ans - Object detection is the process of locating and

classifying objects within images or videos. It involves
identifying and drawing bounding boxes around
objects of interest while also assigning them to
predefined categories or classes, such as "person,"
"car," or "cat." Think of it as teaching a computer to
not only spot items in a picture but also to label them
correctly.

5) Describe object detection history and deep

learning models.

Ans - Object detection has evolved significantly over

the years. Initially, traditional computer vision
techniques relied on handcrafted features and
classifiers to detect objects. However, with the rise of
deep learning, object detection has seen remarkable
advancements.
1. **Traditional Methods**:
- These methods required manual feature
engineering and with complex backgrounds,
variations in scale, and occlusions.
2. **Deep Learning Revolution**:
- The introduction of Convolutional Neural Networks
(CNNs) for improved accuracy and speed.
3. **Single Shot Detectors (SSDs) and You Only Look
Once (YOLO)**:
- They enable real-time detection by directly
predicting bounding boxes and class probabilities
from a single pass through the network.
- YOLO, in particular, gained attention for its
efficiency and speed.
4. **Two-Stage Detectors**:
- Models like Mask R-CNN and Cascade R-CNN
improved upon two-stage detection approaches.
5. **Efficient Object Detection**:
- Models like EfficientDet and YOLOv5 prioritize
accuracy while minimizing computational resources,
making them suitable for various applications,
including edge devices.

Overall, the history of object detection showcases a

transition from handcrafted features to learned
representations through deep learning, leading to
significant improvements in accuracy, speed, and
efficiency.

6) Explain object detection for objects on images with

example.

Ans - Object detection in images involves identifying

and locating various objects within the image.
For example, let's consider an image containing a
street scene with cars, pedestrians, and traffic signs.
Object detection algorithms analyse the entire image,
identify the different objects present (cars,
pedestrians, traffic signs), and draw bounding boxes
around each object while also labelling them with
their respective categories.

This process enables computers to not only recognize

objects but also understand their context within the
image, which is crucial for applications like
autonomous driving, surveillance, and image search.

7) Explain social distancing.

Ans - Social distancing in terms of image processing

and computer vision involves using cameras and
algorithms to monitor and enforce safe distances
between individuals in public spaces.
For example, computer vision algorithms can analyse
video feeds from cameras to detect people and their
positions within a given area. By measuring the
distances between individuals, the system can
identify whether they are adhering to recommended
social distancing guidelines or if they are too close to
each other.
If individuals are found to be too close, the system
can trigger alerts or notifications to remind them to
maintain a safe distance. This technology can be
useful in various settings, such as retail stores,
airports, and public transportation, to help prevent
the spread of contagious diseases.
8) Explain object tracking with example.

Ans - Object tracking involves following the

movement of a specific object in a video sequence
over time.
For example, consider tracking a car in a video of a
busy street. Object tracking algorithms continuously
locate the car's position in each frame of the video,
drawing a bounding box around it. As the car moves,
the algorithm updates the bounding box to reflect its
new position in subsequent frames.
This allows us to monitor the car's trajectory, speed,
and direction of movement. Object tracking is used in
various applications such as surveillance,
autonomous vehicles, and sports analysis.

Unit – 6

1) Define data preparation.

Ans - Data preparation is the process of collecting,

cleaning, organizing, and formatting raw data to
make it suitable for analysis or use in machine
learning algorithms. It involves tasks such as
removing duplicates, handling missing values,
standardizing formats, and transforming data into a
structured format that can be easily processed and
analysed. Think of it as getting your data ready for
the journey ahead, like cleaning and packing your
suitcase before a trip.

2) Explain LineraSVC building with example.

Ans - LinearSVC (Linear Support Vector Classifier) is a

machine learning algorithm used for classification
tasks, particularly in scenarios where the data can be
separated into different classes by a linear boundary.

For example, let's say we have a dataset containing

features of various animals like height, weight, and
diet, and we want to classify them as either
"mammal" or "bird". LinearSVC would learn a linear
boundary in the feature space to separate mammals
from birds.
The algorithm works by finding the optimal
hyperplane that best separates the classes,
maximizing the margin between the classes. It aims
to correctly classify as many instances as possible
while minimizing misclassifications.
Once trained, LinearSVC can predict the class labels
of new data points based on their feature values,
helping classify them into the appropriate categories.

3) Explain RandomForest building with example.

Ans - Random Forest is a popular machine learning

algorithm used for both classification and regression
tasks. It works by constructing multiple decision trees
during training and outputting the class that is the
mode of the classes (classification) or mean
prediction (regression) of the individual trees.
For example, imagine you have a dataset of houses
with features like size, location, and number of
bedrooms, and you want to predict their prices.
Random Forest would build numerous decision trees,
each trained on different subsets of the data and
using random subsets of features.
During prediction, each tree in the forest individually
predicts the price of a house based on its features,
and then the final prediction is determined by
averaging (regression) or voting (classification) the
predictions of all the trees.

This ensemble approach makes Random Forest robust

to overfitting and improves generalization
performance, making it a powerful tool for various
machine learning tasks.

4) Explain prediction of digits on image with example.

Ans - Prediction of digits on an image involves using

machine learning algorithms to recognize and classify
handwritten or printed digits in an image.
For example, suppose you have an image containing
a handwritten digit "5". The prediction process
involves feeding this image into a trained machine
learning model, such as a convolutional neural
network (CNN) or a support vector machine (SVM),
which has learned from thousands of labelled images.
The model analyses the pixel values and features of
the image and predicts that the digit in the image is
most likely "5". This prediction is based on the
patterns and relationships learned from the training
data. The accuracy of the prediction depends on the
quality of the training data and the effectiveness of
the chosen machine learning algorithm.

5) Explain prediction of digits on video with example.

Ans - Prediction of digits on a video involves using

machine learning algorithms to recognize and classify
digits that appear in successive frames of the video.
For example, consider a video stream of handwritten
digits being written on a whiteboard. The prediction
process involves analysing each frame of the video
using a trained machine learning model, such as a
convolutional neural network (CNN) or a recurrent
neural network (RNN).

As the video progresses, the model identifies and

classifies the digits in each frame, providing real-time
predictions of the digits being written. This enables
applications such as digit recognition in live video
streams, interactive educational tools, and gesture-
based interfaces. Ans -

6) Explain image segmentation with example.

Ans - Image segmentation is the process of

partitioning an image into multiple segments or
regions to simplify its representation and facilitate
analysis.
For example, consider an image of a beach scene.
Image segmentation would divide the image into
segments corresponding to different objects or areas
such as sky, water, sand, and people. Each segment
is assigned a unique label or colour, enabling the
identification and analysis of distinct regions within
the image.
This technique is widely used in various applications
such as medical imaging, autonomous driving, and
object detection, where precise delineation of objects
or regions within an image is necessary for further
processing and understanding.

7) Explain colour quantization.

Ans - Colour quantization is the process of reducing

the number of distinct colours in an image while
preserving its visual quality as much as possible.
For example, imagine you have a photograph with
millions of different colours. Colour quantization
would reduce this palette to a smaller set of
representative colours, often with a limited number of
colours (e.g., 256 or fewer).

This reduction helps to decrease the image's file size,

making it easier to store and transmit, while still
maintaining the image's overall appearance to the
human eye. It's like converting a painting from
millions of individual brush strokes to a smaller set of
simplified colours, while still retaining the essence
and feel of the original artwork.

8) Explain vehicle detection with example

Ans - Vehicle detection involves using computer

vision techniques to identify and locate vehicles
within images or videos.
For example, consider a traffic surveillance camera
monitoring a busy intersection. Vehicle detection
algorithms analyse the video feed and identify the
presence and position of cars, trucks, and other
vehicles moving through the intersection.
These algorithms can detect vehicles by recognizing
their distinctive features such as shape, size, and
motion patterns. Vehicle detection is used in various
applications such as traffic management, parking lot
monitoring, and autonomous driving systems to
improve safety and efficiency on the roads.

9) Explain lane detection with example.

Ans - Lane detection is a computer vision technique
used to identify and locate lane markings on roads in
images or videos.
For example, imagine a dashboard camera capturing
footage of a car driving on a highway. Lane detection
algorithms analyse the video frames to detect the
boundaries of lanes by identifying lane markings such
as solid lines, dashed lines, or curbs.
By recognizing the distinct patterns and shapes of
lane markings, these algorithms determine the
position and orientation of lanes relative to the
vehicle. This information is crucial for tasks like lane-
keeping assistance, autonomous driving, and road
monitoring systems, enhancing safety and navigation
on the roads.