Overall Explanation

Overview

With the previous lectures, we looked at different ways to recognize object wihtin an images. Within Image Recognition system, there are some serious problems.

1. The image recognition system cannot locate where the recognized object is within the picture.

2. It has difficulty when it comes to multiple objects within one picture.

Object Detection system allows for solutions for all the downside the Image Recognition system may have. It allows for object detection and object recognition all at the same time, which allows the model to recognize and detect multiple objects at the same time.

The Object Detection system used for our system recieves an image as an input, and outputs a list of coordinates of the detected bounding boxes along with their classes and confidence values. With our follow along example we used SSD-Mobile-v2 model which can detect 91 different classes and was trained on MS COCO dataset.

Having object detection system allows for many different applications ranging from intrusion detections, inventory management, foot traffic analysis and so on.

SSD-Mobilenet-v2

When it comes to most of the object detection models, there are two different categorization:

1. one (single) stage object detectors

2. two (dual) stage object detectors

The difference between the two is that two (dual) stage object detectors first detect where the object is within the image, and then carries out classification only on the detected region. This yields much higher accuracy, but it leads to more computational resources and slower computational time. Example:

• a: Single stage object detector

• b: Dual stage object detector

The SSD-Mobilenet-v2 is a one-stage object detection model. Because the one-stage object detection merges the detection and recognition stage, it is much faster, light and requires less computational power.

Other models for Object Detection

With the jetson-inference examples, other pre-trained detection models are available:

Model	CLI argument	NetworkType enum	Object classes
SSD-Mobilenet-v1	ssd-mobilenet-v1	SSD_MOBILENET_V1	91 (COCO classes)
SSD-Mobilenet-v2	ssd-mobilenet-v2	SSD_MOBILENET_V2	91 (COCO classes)
SSD-Inception-v2	ssd-inception-v2	SSD_INCEPTION_V2	91 (COCO classes)
DetectNet-COCO-Dog	coco-dog	COCO_DOG	dogs
DetectNet-COCO-Bottle	coco-bottle	COCO_BOTTLE	bottles
DetectNet-COCO-Chair	coco-chair	COCO_CHAIR	chair
DetectNet-COCO-Airplane	coco-airplane	COCO_AIRPLANE	airplane
ped-100	pednet	PEDNET	pedestrians
multiped-500	multiped	PEDNET_MULTI	pedestrians, luggage
facenet-120	facenet	FACENET	faces