AM69AX datasheet(270/299 Pages) TI

270 / 299 page

8.4 Accelerators and Coprocessors

8.4.1 GPU

The Graphics Processing Unit (GPU) accelerates 3-dimensional (3D) and 2-dimensional (2D) graphics and

compute applications.

The GPU module is a scalable architecture which efficiently processes a number of different workload

concurrently:

• 3D Graphic Workload, which involves vertex data and pixel data processing for rendering of 3D scenes.

• 2D Graphic Workload, which involves pixel data processing for rendering 2D objects.

• Compute Applications Workload, which involves general purpose data processing.

For more information, see Graphics Accelerator (GPU) section in Processors and Accelerators chapter in the

device TRM.

8.4.2 VPAC

The Vision Pre-processing Accelerator (VPAC) is a set of common vision primitive functions, performing

memory-to-memory (M2M) pixel data processing tasks, such as: color processing and enhancement, noise

filtering, wide dynamic range (WDR) processing, lens distortion correction, pixel remap for dewarping, on-the-fly

scale generation, on-the-fly pyramid generation. The VPAC offloads these common tasks from the main SoC

processors (ARM, DSP, etc.), so these CPUs can be utilized for differentiated high-level algorithms. The VPAC

is designed to support multiple cameras by working in time-multiplexing mode. The VPAC works as front end

to vision processing and prepares frame/scales for further processing by other vision accelerators or processor

cores in the SoC.

For more information, see Vision Pre-processing Accelerator (VPAC) section in Processors and Accelerators

chapter in the device TRM.

8.4.3 DMPAC

The Depth and Motion Perception Accelerator (DMPAC) is a power efficient hardware accelerator that computes

dense stereo depth maps (depth) and dense optical flow vectors (motion) from camera inputs.

The image/video sensor-based environmental perception (also known as scene understanding) is at the core of

many emerging applications in automotive, industrial and consumer electronics. Typically, this involves detection

of all objects in the scene along with their 3D position and motion with regards to the observer or the car by

analyzing one or many related input video streams. Various computer vision algorithms are used to achieve

these tasks.

A very robust method of obtaining the 3D depth from images is to use two cameras in a stereo setup - two

cameras with known relative positions and camera parameters. The two images of the same scene, captured

from two different camera poses/perspectives, are analyzed to find disparities among every pixel positions in the

images. This is known as the Stereo Disparity map. The disparity values of every pixel can be used to obtain the

3D positions of the object/space they belong to via triangulation.

On the other hand, by analyzing two images from a single camera, captured at two different time instances (that

is, two temporal frames in a video), one can determine where each pixel in a past frame moved to in the future

frame. This is known as the Optical Flow vector. The flow vectors for each pixel position can be used to obtain

3D structure of the scene, identify moving objects and determine their relative speed and direction of motion.

The DMPAC is dedicated to the aforesaid image processing tasks. The stereo and optical flow processing is

partitioned into two top level sub-blocks: the Dense Optical Flow (DOF) engine and the Stereo Disparity Engine

(SDE). The DOF and SDE blocks share a common shared local memory, DMA, external messaging and control

infrastructure.

For more information, see Depth and Motion Perception Accelerator (DMPAC) section in Processors and

Accelerators chapter in the device TRM.

AM69A

SPRSP92 – FEBRUARY 2023

www.ti.com

270

Submit Document Feedback

Product Folder Links: AM69A

AM69AX Datasheet(PDF) 270 Page - Texas Instruments

AM69AX Datasheet(HTML) 270 Page - Texas Instruments

Html Pages

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 ...More

Datasheet Download

Link URL