|
YOLO (YOU ONLY LOOK ONCE) Making Object detection work in Medical Imaging on Convolution detection System
|
|
| Author:
|
1SYED INTHIYAZ, SK HASANE AHAMMAD, A SAI KRISHNA, V BHARGAVI, D GOVARDHAN, V RAJESH
|
| Abstract:
|
We introduce the emerging and the unified architecture in the field of object detection techniques that have made different approach resulting the other phase of detection and classification. YOLO medical image is the model which computes network as relapse of input image and builds individual bounding boxes for every associated class object with more accuracy. As the model convolves itself into a lone neural network, jumps straight to the image pixels to bounding box coordinates and object classes. A uni-neural network reproduces the detection boxes and classify probabilities straight from full images in single time instance. Being a single network consisted pipeline, it can be more stressed on the detection performance rather than the prediction boxes. YOLO is fast as a swift compared to other systems. Its architecture allows to pre-process the images for detection at 45 frames per second when experimented. We also trained the model with more weights and gaining the maximum number of objects to detect, which in turn produces as accurate as three times faster than SSD. Compared to world of detection systems, YOLO medical image system leads to more localization errors but is less probable in predicting the background junk. YOLO outperforms the detection methods in a smoothing way from others like Medical image DPM, SSD, CNN and R-CNN.
|
| Keyword:
|
Computer vision, object detection, bounding box, class prediction, convolutional neural networks (CNN), feature extraction, Multi-labelling, localization.
|
| EOI:
|
-
|
| DOI:
|
https://doi.org/10.31838/ijpr/2020.12.02.0003
|
| Download:
|
Request For Article
|
|
|