Question

10. (5 pts) Describe the features of Faster R-CNN, Region-based Fully Convolutional Network (R-FCN) and Single-shot Multibox Detector (SSD). Compare the three object detectors in terms of accuracy and speed.

Answer 1

Answer: The mean average precision (mAP) values of the three detectors are as following.

• Faster R-CNN using inception resnet is having approximately 34 mAP.
• R-FCN is having approximately 13 mAP.
• Single-shot Multibox Detector (SSD) is having approximately 19 mAP.

From the above values, precision value for faster R-CNN is greater than the remaining two values. So, faster R-CNN is more accurate even it takes more time. SSD and R-FCN are less accurate but they are faster than R-CNN.

​​