Week 3— Object Detection for Blind People

Hello everyone. This is our third week. This week we will write about why we selected the YOLO algorithm and alternatives of YOLO. Also, we write about the difference between object detection and object classification.

When we search about object detection algorithms we find these algorithms:

1-Region-based Convolutional Neural Networks(R-CNN)

2-Fast R-CNN

3-Faster R-CNN

4-YOLO(You Only Look Once)

5-Single Shot Detector(SSD)

After finding these algorithms we start to research which one is the best and which one is faster.

Real-Time Systems on PASCAL VOC 2007. Comparison of speeds and performances for models trained with the 2007 and 2012 PASCAL VOC datasets.

Detection frameworks on PASCAL VOC 2007. YOLOv2 is faster and more accurate than prior detection methods. It can also run at different resolutions for an easy tradeoff between speed and accuracy. Each YOLOv2 entry is actually the same trained model with the same weights, just evaluated at a different size. All timing information is on a Geforce GTX Titan X (original, not Pascal model).

As we can see from the tables YOLO is faster than others. Also now there is a better version of the YOLO (YOLOv3).

For our project, we use object detection. Object detection can be confused with object classification. The difference between object detection and object classification is that in classification tasks, the classifier outputs the class probability whereas, in object detection tasks, the detector outputs the bounding box coordinates and the predicted classes.

A demonstration from the YOLOv3:

After detecting the object we need to in a frame, the next step is to obtain the depth or distance of the detected object from the user. We are going to research about it.

Thank you for reading.