Page 326 - AI Computer 10

2. You are tasked with developing a computer vision system for a self-driving car company. The system needs
to accurately detect and classify various objects on the road to ensure safe navigati on. Imagine you're
working on improving the object detecti on algorithm for the self- driving car's computer vision system.
During testi ng, you noti ce that the system occasionally misclassifi es pedestrians as cyclists, especially in
low-light conditi ons.
a. How would you approach addressing this issue?
b. What steps would you take to enhance the accuracy of pedestrian detecti on while ensuring the
system's overall performance and reliability on the road?

Ans:
a. The issue can be addresses by improving the training data by incorporati ng diverse lighti ng scenarios,
refi ning the object detecti on model with features that highlight pedestrian-specifi c att ributes such as
pose and stance, and integrati ng additi onal sensor data for more robust detecti on.
b. Some steps to enhance the overall performance and reliability of the system are:
(i) Gather a large, varied dataset comprising of images under diff erent lighti ng conditi ons during
diff erent periods of the day and night, weather variati ons, clothing styles, and camera angles to
train the model on a wider range of pedestrian appearances.
(ii) Label images to emphasise features that disti nguish pedestrians from cyclists, such as leg posture,
arm movement, and clothing details.
3. A city’s transportati on department wants to reduce traffi c congesti on using AI. They install smart traffi c
cameras at busy intersecti ons to analyse vehicle movement and adjust traffi c signals dynamically. The
system uses computer vision to detect vehicle density and classify vehicles (cars, buses, motorcycles).

Answer the following questi ons based on the case study.
a. How does computer vision help in real-ti me traffi c management?
b. If a no-code AI tool like Teacable Machine is used for this applicati on, what steps would be involved
in training the model?

Ans:
a. Computer vision processes live traffi c footage to count vehicles, detect congesti on patt erns, and
adjust traffi c signals accordingly to improve fl ow.
b. Step 1: Upload labeled images of diff erent vehicle types.
Step 2: Train the model using classifi cati on techniques.

Step 3: Test and fi ne-tune the model based on accuracy metrics.
Step 4: Deploy the model to process live video feeds.

AI Assessment Zone

A. Tick () the correct answer.

1. Which task in Computer Vision involves both object localization and object classification?
a. Image classifi cati on b. Object Recogniti on

c. Object Detecti on d. Semanti c Segmentati on

192
192