Uploaded by Surya Kotamraja

open lab report 1

advertisement
Open Lab Report
Autonomous Underwater Vehicle (AUV) using YOLO V5 Algorithm
ByBoomu Nagendra Reddy
BL.EN. U4EAC21015
Sreekar Prabhas C H
BL.EN. U4EAC21070
Surya Vamsi Kotamaraja
BL.EN. U4EAC21072
Abstract:
Underwater object detection plays a pivotal role in various applications ranging
from marine biology research to underwater exploration and surveillance. This
paper presents a novel approach to underwater object detection using the
YOLOV5 algorithm. The proposed underwater object detection model leverages
the capabilities to efficiently and accurately identify and localize objects in
underwater imagery. We used the aqua pre-train dataset for training and
detecting distinct fish species in different water bodies and achieved the highest
mean average precision value of 0.861 for the jellyfish class. This proposed
method is helpful in identifying the debris or structures located in the depths of
oceans and dealing with underwater calamities.
Introduction:
The world's oceans remain one of the most mysterious and least explored
frontiers on our planet. Beneath the surface of the water lies an intricate and
diverse ecosystem, teeming with marine life and concealed treasures.
Understanding and monitoring this hidden world is crucial for a wide range of
applications, including environmental conservation, marine resource
management, underwater archaeology, and offshore infrastructure
maintenance. Deep learning innovations have transformed computer vision in
recent years, providing a potent tool for underwater object detection and
recognition. This research paper delves into the realm of underwater object
detection, focusing on the application of the YOLOv5 (You Only Look Once
version deep learning model.
A cutting-edge real-time object recognition system with a reputation for speed
and accuracy, YOLOv5 is a strong contender for solving the particular issues
presented by underwater environments. The study of deep seas and oceans
often includes lasers to understand submerged objects. Artificial intelligence
and computing technology have advanced quickly, gaining higher accuracy and
processing speeds. Deep learning has vast real-time applications in different
domains like speech recognition, face recognition, handwritten text
recognition, automated image descriptions, and many more.
YOLO V5:
YOLOv5 stands out in the realm of object detection models for its exceptional
balance of speed and accuracy. It can identify and classify objects in videos and
images in real time, making it a powerful tool for various applications. Unlike
traditional object detectors that require multiple processing stages to predict
bounding boxes and class probabilities, YOLOv5 performs these tasks in a
single step, significantly streamlining the detection process. This efficiency
makes it suitable for real-time tasks where fast response times are crucial, such
as self-driving cars, autonomous robots, and drone surveillance systems.
YOLOv5 offers users the flexibility to choose from different model sizes (s, m,
l, x) that prioritize speed or precision depending on the specific needs of the
application. For instance, if real-time processing is critical for a video
surveillance system, a smaller, faster model (e.g., YOLOv5s) might be
preferable to ensure minimal delays in identifying potential security threats.
Conversely, if the highest possible accuracy is desired for a medical image
analysis application, a larger, more precise model (e.g., YOLOv5x) can be
selected to provide more reliable diagnoses.
Finally, YOLOv5 is an open-source project, which means it's freely available
for anyone to use and modify. This open-source nature fosters a collaborative
development community that continuously improves the model's capabilities.
Additionally, it boasts a user-friendly design with well-documented resources,
making it accessible to developers of all experience levels. Those who want to
delve deeper can explore the extensive documentation to gain a comprehensive
understanding of the model's architecture, training processes, and potential
applications. Alternatively, users can leverage pre-trained models for quick
deployment in tasks like object recognition in videos or traffic monitoring.
YOLOv5's ability to be fine-tuned on custom data to detect new object classes
makes it a highly versatile tool, allowing developers to adapt it to specific needs
and domains.
Code:
!pip install ultralytics
importos
importcv2
importnumpy asnp
importmatplotlib.pyplot asplt
fromultralytics importYOLO
fromIPython.display importImage,display
root_dir = '/content/drive/MyDrive/archive (4)/aquarium_pretrain'
train_img_dir = '/content/drive/MyDrive/archive
(4)/aquarium_pretrain/train/images/IMG_2274_jpeg_jpg.rf.2f319e949748145
fb22dcb52bb325a0c.jpg'
test_img_dir = '/content/drive/MyDrive/archive
(4)/aquarium_pretrain/test/images/IMG_2289_jpeg_jpg.rf.fe2a7a149e7b11f2
313f5a7b30386e85.jpg'
train_img = cv2.imread(os.path.join(train_img_dir),cv2.IMREAD_COLOR)
test_img = cv2.imread(os.path.join(test_img_dir),cv2.IMREAD_COLOR)
plt.figure(figsize=(10, 5))
plt.subplot(1, 2, 1)
plt.imshow(train_img)
plt.axis('off')
plt.subplot(1, 2, 2)
plt.imshow(test_img)
plt.axis('off')
plt.show()
model = YOLO('yolov8n.yaml')
results = model.train(data = os.path.join(root_dir, 'data.yaml'),epochs
= 5)
detection_pipeline(test_imgs)
Data-set-
Output:
Results:
Our proposed model utilizes the YOLOV5 object detection model to identify
objects in different marine and sea bodies. The environment underwater is very
unpredictable and has challenging situations. But our model can overcome all
these tough conditions and analyze the images to perform feature extraction and
then object detection and classification.
The algorithm's capacity to manage numerous item classes concurrently makes
it possible to detect a variety of underwater objects in a single pass. With this
model, we have been able to identify and segregate the underwater species into
seven different categories fish, jellyfish, starfish, puffins, penguins, stingrays,
and sharks.
Reference:
K. Meghana, R. Vandana, P. V. Kumar, P. Swetha and M. A. Jabbar, "Detection
of underwater objects using Deep Learning and You Look Only Once
Algorithm," 2023 IEEE 3rd Mysore Sub Section International Conference
(MysuruCon), HASSAN, India, 2023, pp. 1-7, doi:
10.1109/MysuruCon59703.2023.10396887.
N. Reddy Nandyala and R. Kumar Sanodiya, "Underwater Object Detection
Using Synthetic Data," 2023 11th International Symposium on Electronic
Systems Devices and Computing (ESDC), Sri City, India, 2023, pp. 1-6, doi:
10.1109/ESDC56251.2023.10149870.
A. Balaji, Y. S, K. CK, N. R, G. Dooly and S. Dhanalakshmi, "Deep WaveNetbased YOLO V5 for Underwater Object Detection," OCEANS 2023 - Limerick,
Limerick, Ireland, 2023, pp. 1-5, doi:
10.1109/OCEANSLimerick52467.2023.10244645.
S. Raavi, P. B. Chandu and S. T, "Automated Recognition of Underwater
Objects using Deep Learning," 2023 7th International Conference on Trends in
Electronics and Informatics (ICOEI), Tirunelveli, India, 2023, pp. 1055-1059,
doi: 10.1109/ICOEI56765.2023.10125839.
Download