You are on page 1of 3

Lab 11

 Object: To Learn the basics of computer vision modeling by applying a typical


workflow—tracking-by- detection—to a video of turtles crawling toward the sea and learn
about the role of features in computer vision, how to label data, train an object detector, and
track wildlife in the video.

 Introduction:
 Computer Vision:
Artificial intelligence (AI)'s field of computer vision enables computers and systems to extract useful
information from digital photos, videos, and other visual inputs and to conduct actions or offer
recommendations in response to that information.
If AI gives computers the ability to think, computer vision gives them the ability to see, observe,
and comprehend. Human vision has an advantage over computer vision in that it has been around
longer. With a lifetime of context, human sight has the advantage of learning how to distinguish
between things, determine their distance from the viewer, determine whether they are moving, and
determine whether an image is correct.

 Computer Vision On-Ramp:

 Introduction to course:
Computer vision has developed into a potent tool for the study of visual data, enabling us to extract
useful information from images and movies. This lab focuses on using a tracking-by-detection
mechanism to track turtles in a movie using computer vision techniques. Understanding the
principles of computer vision modeling and overcoming the challenges involved in wildlife
monitoring are the key objectives.
 Features & their importance:
Features are crucial in computer vision tasks because they enable the identification and tracking
of objects of interest. With the help of feature-based analysis of the video frames, turtles are
located and monitored in this lab. Features are extracted from each frame and compared in
order to identify and follow the turtles' movement.
 Importance of labels:
The training data must be labelled with bounding boxes that correctly represent the locations of
the turtles in order to train an item detector. The turtles are methodically boxed in on every frame
of the film for annotation purposes. Accurate labeling directly affects the model's performance,
hence additional care is used to ensure it.
 Object detector training:
After the data has been labelled, an object detector is trained using it. To understand the visual traits
of turtles, the team uses a well-liked object identification method, such as Faster R-CNN or YOLO.
The labelled data are fed into the model during training, and the model's parameters are optimized
along with the detection performance repeatedly.
 Tracking-by-detection:
After training the object detector, the tracking-by-detection methodology is utilized to locate turtles
in the movie. Turtles are recognized in each frame using the object detector, and the bounding box
coordinates are recorded. In order to track particular turtles throughout the film, the coordinates are
then utilized to generate trajectories by connecting them across frames.
 Output:
 The results of the tracking-by-detection method are presented, showcasing its capability
to precisely track turtles in the video. The limitations and challenges of the method are
discussed, including instances of occlusion- or illumination-related detection failures.
 The object detector's accuracy and robustness are also evaluated using precision and
recall metric.

 Conclusion:
The approach and findings are reiterated in the lab report's conclusion. Using computer vision
modeling, particularly tracking-by-detection, turtles in the movie were successfully tracked. The
paper emphasizes the importance of precise data labeling, the value of features in computer vision,
as well as the limitations and challenges of utilizing computer vision to monitor animals.

You might also like