Scribd
Upload
19 views2 pages

Summary PDF

The document outlines the performance evaluation of two models: a YOLO-based object detection model for drone detection and a segmentation model for football player images. The YOLO model achieved a precision of 0.904 and a mean Average Precision (mAP) of 0.862, while the segmentation model reported a Dice Score of 0.84 and pixel accuracy of 92.5%. Both models utilized various techniques such as data augmentation and hyperparameter tuning to enhance performance.

Uploaded by

poojaexplore1
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
19 views2 pages

Summary PDF

The document outlines the performance evaluation of two models: a YOLO-based object detection model for drone detection and a segmentation model for football player images. The YOLO model achieved a precision of 0.904 and a mean Average Precision (mAP) of 0.862, while the segmentation model reported a Dice Score of 0.84 and pixel accuracy of 92.5%. Both models utilized various techniques such as data augmentation and hyperparameter tuning to enhance performance.

Uploaded by

poojaexplore1
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd

Task Model Measures take to improve Evaluation metrics

Architecture model performance


Object Detection model = Data Precision (P) = 0.904
(Drone detection) YOLO("yolov8x. augumentation(augument=T
pt") rue) Recall (R) = 0.742
Train-1012 .jpg
images + labels #extra large mAP@IOU 0.5= 0.862
variant of Hyperparameter tuning
Test-347 files .jpg yolov8,trained
on COCO, 80 With extra-large variant,used
Valid-347 files .jpg class detect
!yolo task=detect mode=train
model=yolov8x.pt
data=data.yaml epochs=20
imgsz=480

Distribution of IoU:
The IoU histogram shows a
concentration of high IoU values
(0.7-1.0), indicating that most
predictions closely align with
ground truth boxes.

Error Analysis: Some detections


have low IoU values (~0.0-0.2),
suggesting occasional false
positives or poor localization.
Image (512, 512, 3) Dropout layers for IoU = 0.78
Segmentation Conv2D, regularization, Skip connections
(Football players MaxPooling, for better feature preservation Dice Score = 0.84
image Conv2DTransp
segmentation) ose, Dropout, Pixel Accuracy = 92.5%
Concatenation
training 368
validation: 92
test:52

You might also like