Object Detection

Morpho Deep Detector™

“Morpho Deep Detector” is a general-purpose API library for AI-based object detection tasks. It is built on the latest version of a fast inference engine “SoftNeuro®” and realizes highly accurate object detection at high speed and low power consumption.

Use Case

Function or Purpose

Features

Built-in fast AI inference engine "SoftNeuro®" realizes high speed and low power consumption in accordance with hardware
Operates on a variety of platforms, including PCs, embedded devices, and server applications
AI model capable of detecting 80 items is included
The learning model can be switched among five levels for pursuing detection accuracy or processing speed, depending on the application.

Product Details

In AI-based applications, object detection tasks are widely used as a fundamental technology. Examples of applications include pedestrians and vehicles detections in automated driving (AD) and advanced driver assistance systems (ADAS) which use in-vehicle cameras in the automotive field. In-vehicle cameras are also used for the recognition of occupants and objects. In the field of surveillance cameras, they are widely used to recognize people and their conditions.

Morpho has updated its object detection engine “Morpho Deep Detector™” to enable its use in such a wide range of applications.

(1) Updated “SoftNeuro®”, a built-in AI inference engine.
The latest version of “SoftNeuro”, a fast inference engine, is built into the “Morpho Deep Detector”. It supports a variety of hardware (CPU, GPU, and various AI accelerators). By maximizing the use of hardware resources, it realizes highly accurate object detection at high speed while keeping power consumption low. Operating environments such as x86/x64/ARM CPUs, GPUs, and various AI accelerators are available.

AI inference engine “SoftNeuro”: https://www.morphoinc.com/en/technology/sie

(2) AI model capable of detecting 80 items is included as a standard model
The learning model included in the product has been redesigned, and it is now capable of detecting 80 items*. In addition, the learning model can now be switched among five levels for pursuing detection accuracy or processing speed, depending on the application. This provides more options to users and allows more freedom in application development.

Person		Road-related		Tableware		60	bed
1	person	20	traffic light	40	bottle	61	dining table
Vehicle		21	fire hydrant	41	wine glass	62	toilet
2	bicycle	22	stop sign	42	cup	63	sink
3	car	23	parking meter	43	fork	Home Appliance
4	motorcycle	24	bench	44	knife	64	laptop
5	airplane	Portable Goods		45	spoon	65	mouse
6	bus	25	backpack	46	bowl	66	remote
7	train	26	umbrella	Food		67	keyboard
8	truck	27	handbag	47	banana	68	cell phone
9	boat	28	tie	48	apple	69	microwave
Animal		29	suitcase	49	sandwich	70	oven
10	bird	Sporting Goods		50	orange	71	toaster
11	cat	30	frisbee	51	broccoli	72	tv
12	dog	31	skis	52	carrot	73	refrigerator
13	horse	32	snowboard	53	hot dog	General Goods
14	sheep	33	sports ball	54	pizza	74	book
15	cow	34	kite	55	donut	75	clock
16	elephant	35	baseball bat	56	cake	76	vase
17	bear	36	baseball glove	Furniture		77	scissors
18	zebra	37	skateboard	57	chair	78	teddy bear
19	giraffe	38	surfboard	58	couch	79	hair drier
		39	tennis racket	59	potted plant	80	toothbrush

The Format to Provide the Product

Library files and technical documentation for application development are provided to suit your operating environment. Please contact our representative for details.

Contact