YOLO -- You only look 10647 times

Abstract

With this work we are explaining the "You Only Look Once" (YOLO) single-stage object detection approach as a parallel classification of 10647 fixed region proposals. We support this view by showing that each of YOLOs output pixel is attentive to a specific sub-region of previous layers, comparable to a local region proposal. This understanding reduces the conceptual gap between YOLO-like singlestage object detection models, RCNN-like two-stage region proposal based models, and ResNet-like image classification models. In addition, we created interactive exploration tools for a better visual understanding of the YOLO information processing streams: https://limchr.github.io/yolo_visualization

Faces: AI Blitz XIII Solutions

“…See more details on our implementation in the Age Prediction notebook [5]. We have achieved good results in age prediction using the You Only Look Once (YOLO) architecture [9,10,11]. Although YOLO is an object detection algorithm, we mainly utilized its class prediction capabilities and abandoned the bounding box prediction functionality.…”

Face Generation and Editing With StyleGAN: A Survey

“…Expressions and emotions from the target image are kept, while the facial identity is swapped. First, it collects a dataset of faces of two people, A and B, using an object detection method [102]. Secondly, it trains two auto-encoders E A , E B to encode and two decoders DC A , DC B to reconstruct the faces of A and B respectively.…”

A Bus Crowdedness Sensing System Using Deep-Learning Based Object Detection

“…During training, 200 images were assigned to the test dataset In the training process of the object detection, we need to define 9 anchor frames to accelerate the model training and improve the detection accuracy. The anchor frames are predefined bounding box patterns used by YOLOv5 to delineate regions for object candidates [19]. Each frame (x, y) indicates the width and height of the target size prediction bounding box.…”

“…These frames were selected by using kmeans to cluster the labels in the training set. Please refer to [19] for more details regarding the selection of anchor frames. In this study, the following anchor frames are used.…”