?? Comparison R-CNN, Fast R-CNN & Faster R-CNN (Object Detection)

Trying to show comparison between various R-CNN algorithms used for any Object Detection Problem.

?? Table below provides a good overview of results of different networks I will discuss in coming time

?? Problem with R-CNN:

• It uses Selective Search Algorithm to find Regions of Interest which is a slow and time consuming process
• Each image needs to classify?(~2000)?region proposals. So it takes a lot of time to train network, assume for 3000 images total number of features?for CNN will become?6,000,000 i.e. (3000 x 2000)
• It requires?(~49)?seconds to detect objects in an image on GPU
• To store feature map of region proposal, lots of Disk space is also required

Fast-R CNN was developed in order to overcome above mentioned bottleneck of RCNN

?? Problem with Fast R-CNN:

• Fast R-CNN still uses Selective Search Algorithm which is slow and a time-consuming process
• It takes around 2 seconds per image to detect objects, which sometimes does not work properly with large real-life datasets
• Most of time taken by Fast R-CNN during detection is a selective search region proposal generation algorithm. So it is bottleneck of this architecture which was dealt with in Faster R-CNN

Keep in you head:

Instead of Selective Search algorithm, Faster R-CNN uses RPN (Region Proposal Network) to select best ROIs automatically to be passed for ROI Pooling

Do you want me to simplify CNN, R-CNN, Fast R-CNN, Faster R-CNN??