
A4 Tri-fold brochure for FastVis FPGA Camera.

Full Bleed Size
282 x 219 mm
3331 x 2588 pixels
Document Trim Size
279 x 216 mm
3295 x 2551 pixels


(Templates/images attached)





Text content:

-----------------------------------------------------------------------------

FastVis FPGA
Camera

Information, not 
just data

-----------------------------------------------------------------------------

2 Megapixel, 12BPP
340 frames/sec
150k LUTs FPGA
USB3
1.5Gb RAM
Integrated IMU
Small, lightweight
Low power
Open source drivers
FastXForm 
  C-to-camera service

Web: fastvis.com
Email: info@fastvis.com


-----------------------------------------------------------------------------

Designed for Computer Vision

Vision has long been recognised as the most 
powerful sensing modality. Computer vision is 
all about extracting meaningful information 
from the abundance of data provided by video 
cameras.  The sheer amount of raw data 
presents challenges: processing the data is 
difficult and it is non-trivial just to transfer 
the data to a computer.

The FastVis FPGA camera addresses 
technological challenges in computer vision 
by combining a powerful FPGA directly with 
a 2 megapixel camera.  


The FastVis camera can be used for all types 
of computer vision algorithms.  E.g.
  * edge, feature detection
  * optical flow
  * object tracking & control
  * image and video filtering
  * downsampling
  * upsampling




-----------------------------------------------------------------------------

Specifications:


Imaging Sensor
Resolution		2048x1088 pixels, 12BPP
Frame Rate 		340 Hz 
Shutter 		Global 
Colour 			Bayer RGB 
Sensitivity 		5.56 V/lux.s
Default dynamic range 	60 dB
High dynamic range 	Piecewise linear/
     	     		variable sensitivity
Optical format 		2/3 inch

FPGA
Logic Cells		149,000
Cell Architecture 	4 input lookup table&FF
18x18 Multipliers 	320
Internal RAM 		7Mbits
External RAM 		1.5Gbit
External RAM bandwidth 	>10Gbps (goal is 15Gbps)

Connections
USB			USB 3.0
USB throughput		Up to 3.0Gbps
Inter-camera  		3Gbps links
GPIO			16 bi-directional GPIO

IMU
Gyroscope		3-axis
Gyro range 		+/- 250 to +/- 2000 deg/s
Gyro sensisitivity 	0.007 deg/s to 0.06 deg/s
Gyro sample rate 	up to 8kHz
Accelerometer 		3-axis
Accel range 		+/- 2g to +/- 16g
Accel sensisitivity 	6e-5g to 5e-4g
Accel sample rate 	up to 1kHz

Misc
Lens Mount		C-mount
Camera Mount 		Standard camera (1/4" 20TPI)
Dimensions 		50.8mm x 50.8mm x 42mm (excl. lens)
Weight 			85-90g (excl. lens)
Power 			5V, 1A


[as a table, perhaps with alternate rows
having shaded backgrounds?]


-----------------------------------------------------------------------------
FastXForm
FastXForm is an online service that provides 
translation from a subset of C to a bitstream that 
will run on the FastVis camera.
* Enables rapid development. Develop and debug 
    on standard PC using FastVis host library
* Easy to use, well known language.
    C is much easier than Verilog/VHDL.
* FastXForm takes care of scheduling and pipelining
* Integrated with device drivers which means you 
    can focus on the computer vision algorithm
* Modules written as a per-pixel function call
* fv_get_pix function allows access to pixels
    offset by row, column and even frame
* Optimises constant multiplication, provides
   square-root and division
* Low cost

Example of 5x5 Gaussian Blur:

uint16_t gaussian_kernel[5] = {
  1,  4,  6,  4, 1
};
void gaussian_5x5_x(fv_stream_t *stream) {
  uint64_t ret = 0;
  int i;
  for (i = -2; i < 3; i++) 
    ret += gaussian_kernel[i] * 
             fv_get_pix(stream, i, 0, 0);
  fv_put_pix(ret / 16);
}
void gaussian_5x5_y(fv_stream_t *stream) {
  uint64_t ret = 0;
  int i;
  for (i = -2; i < 3; i++) 
    ret += gaussian_kernel[i] * 
             fv_get_pix(stream, 0, i, 0);
  fv_put_pix(ret / 16);
}

[use a mono-spaced font for the code above to preserve 
formatting]
