Object Count Video
Please note that the object count is not being
held out as fully accurate. There will be times when it will get it
wrong even if it is set up correctly. However it is a good
backup check and gives a fair amount of accuracy. The accuracy
increases when there is a very good controlled environment.
Our general picture recognition software is
designed to find pictures and objects that are similar to a sample object
within a sample picture. This is the main way in which the software
runs. In short the software is a very powerful system for finding
pictures and then tagging those pictures so they can easily and importantly
quickly be found in the future using our specially designed picture
retrieval system so that any picture on a computer, network drive or
any other media that is connected to the computer showing a windows drive can be quickly found.
This is the main side of the software but sometimes under a controlled
environment a business may wish to count the number of objects and run a
check to make sure that it is the correct amount specified or to
ignore the amount specified.
The applications for detecting an object
is huge but we have kept with our principal that any object shape as long as
the colours in the object are unique from the background colour the
software will find any shape object hopefully.
actually photo being taken for best results requires the same background to
importantly, always to be the same colour for the object count to work
well. If you are just trying to count the amount of any object in a photo
then if the background changes it may not provide best results.
2) The object colour
and the background colour must be different for example black background
white foreground objects.
3) The objects must be a similar size
If you can control the pictures to this degree then
our software can check the results and the level of accuracy should be quite
high. If you already have a video system that can produce stills by
clicking a button again this can be used because the resolution can be very
4) All the pictures taken must be sent to a folder. That
will then be the folder that our software will analyze each pictures from. A
different batch of picture for example could be put in different folders.
5) For best results the pictures should make the objects as large as
possible but the size should be then kept the same for each picture taken.
That really is about it if this can be controlled then the software can
count to a good level of accuracy thousands of pictures and count and check
each picture for the required amount of objects in that photo..
software can also produce an analysis of the results so that the user can check
and view how well the software is working on say the first 10 images of
maybe 5000 images to check that the software is calibrated correctly for
each object counted.
Additional things the user can do to improve the
accuracy of the checks is manually set the height and width and how much gap
is between objects.
This is basically a counting object function
using our picture recognition technology in a slightly different way.
The more development we do the more smaller the objects that are being
counted can be. This is an interesting new function added to our picture
recognition system that allows controlled object within controlled pictures
to be counted with a high amount of accuracy.
There is also a
read out provided showing the number of objects that should be listed or
ignored. The weighting of each object counted and the colour depth of
each object. This information is provided so the user can analyze
results and improve the accuracy of the first several counts by changing
parameter within the software if required. This will help insure best
accuracy for large batch counts of objects.
The better the controlled
environment the better the results. If lighting is correctly set
within the place the photo is actually being taken and the camera is
connected directly to a computer at a set location away from the objects
then results will always be better. The more control you have for example
using a tripod or making sure that the video camera that can also take
stills is at a fixed location. Then results will always be improved.
But if time is taken to set up a controlled environment then the
results and accuracy levels should be very high indeed. The better the
picture for example the less spacing is required between objects or the
smaller the objects can actually be.
The video below gives a good
idea of how the object count system actually works. To challenge our
software we do not control our environment we just take photos using things
like mobile phones. This really does help develop how clever the
software needs to be therefore when images are taken in a controlled
environment results should be very high indeed.
Our New Object Count Video