GE competition to catch cheats

GE competition to catch cheats

Post by Gafoo » Thu, 20 Oct 2005 11:47:38


"Bend it like Murali" competition organized by GE
http://www.expressindia.com/cricket/fulliestory.php?content_id=80243
 
 
 

GE competition to catch cheats

Post by India Fa » Thu, 20 Oct 2005 12:53:05

Quote:

> "Bend it like Murali" competition organized by GE
> http://www.expressindia.com/cricket/fulliestory.php?content_id=80243

That'd be awesome if this thing comes through!

I'm pretty sure this'll benefit Murali more than anybody else.  I'd be
more curious to see how McGrath fares given the not too recent
discovery that he (along with pretty much everyone else) chucks too
as per the old rules.

 
 
 

GE competition to catch cheats

Post by Colin Kynoc » Thu, 20 Oct 2005 13:17:01


much thought and consideration decided that the following would
improve the lives of those that read it:

Quote:
>"Bend it like Murali" competition organized by GE
>http://www.expressindia.com/cricket/fulliestory.php?content_id=80243

Nice idea, but basically useless to tell if a bowler chucks or not as
it appears they will only be using one camera, and therefore it will
only calculate the angle in two dimensions rather that the three
required

Colin Kynoch

 
 
 

GE competition to catch cheats

Post by Gafoo » Thu, 20 Oct 2005 13:41:14

Quote:


>> "Bend it like Murali" competition organized by GE
>> http://www.expressindia.com/cricket/fulliestory.php?content_id=80243

> That'd be awesome if this thing comes through!

> I'm pretty sure this'll benefit Murali more than anybody else.  I'd be
> more curious to see how McGrath fares given the not too recent
> discovery that he (along with pretty much everyone else) chucks too
> as per the old rules.

You don't need any slow motion or replays or angles or competition
to figure out that Murali chucks & chucks bigtime. There is no
comparison. Even at the school & college level people who chuck
as obviously as Murali won't be allowed to play.
 
 
 

GE competition to catch cheats

Post by driz » Fri, 21 Oct 2005 06:46:09

I agree with Colin ... just a lateral image is not sufficient to
conclude the bending.
Besides, the human skin color is not the same .. so the contraption
needs to address that also.

To me it seems like a BS undergrad project.

 
 
 

GE competition to catch cheats

Post by Bounce I » Fri, 21 Oct 2005 06:56:52


Quote:
>I agree with Colin ... just a lateral image is not sufficient to
> conclude the bending.
> Besides, the human skin color is not the same .. so the contraption
> needs to address that also.

> To me it seems like a BS undergrad project.

So you understand more about this than the people working on the project.
And your profession is???
 
 
 

GE competition to catch cheats

Post by driz » Fri, 21 Oct 2005 08:46:27

Why is my profession important?
Anyone who has played cricket and has an engineering background can
tell that a solution of determining the flexing of one's arm 'at the
point of delivery' based on a 2-D model is BS.
It sounds more like a joke when there are camera phones used in the
model, and when color noise can degrade the results. Are they going to
have sight-screens either side of the bowler now? What about the top
view? Are they going to require pitches of a certain color?
What about the capture rate of the camera? How accurate will it be when
clicking from 100m away, even if it does 100 fps? How will this system
measure when the ball is actually released from the fingers?

May be they came up with a novel algorithm for motion detection. And it
may be a good algorithm for detecting the angle between two co-planar
lines that are in motion. But the human body parts' movement is much
more complex a process to be measured by a simple model developed in 2
days.

Calling it a BS may be too harsh, but I am very skeptical about it.

 
 
 

GE competition to catch cheats

Post by driz » Fri, 21 Oct 2005 08:48:11

Can Murali or his lawyers sue the Shastra organizing board?

 
 
 

GE competition to catch cheats

Post by dodo » Fri, 21 Oct 2005 09:04:25

Quote:

> Why is my profession important?
> Anyone who has played cricket and has an engineering background can
> tell that a solution of determining the flexing of one's arm 'at the
> point of delivery' based on a 2-D model is BS.
> It sounds more like a joke when there are camera phones used in the
> model, and when color noise can degrade the results. Are they going to
> have sight-screens either side of the bowler now? What about the top
> view? Are they going to require pitches of a certain color?
> What about the capture rate of the camera? How accurate will it be when
> clicking from 100m away, even if it does 100 fps? How will this system
> measure when the ball is actually released from the fingers?

> May be they came up with a novel algorithm for motion detection. And it
> may be a good algorithm for detecting the angle between two co-planar
> lines that are in motion. But the human body parts' movement is much
> more complex a process to be measured by a simple model developed in 2
> days.

> Calling it a BS may be too harsh, but I am very skeptical about it.

But it is not.

Pataudi and Milburn had one eye and played a 3D ball reasonably well.

 You can close one eye and tell a lot about the three-dimensional
world. Why is that?
1) you have models of how things should be/are and so you can
unambiguously say how their 2D projections will look like?
2) you see things in relationship to each other
3) You know how light shines and reflects of objects, and shadows and
highlights tell you a lot.

In fact the fact you have two eyes gives you almost no three
dimensional advantage for small objects at distances more than twenty
meters away. It is these other ways of perceiving three-D that let you
see that.

 
 
 

GE competition to catch cheats

Post by amukho » Fri, 21 Oct 2005 09:15:08

Quote:

>  You can close one eye and tell a lot about the three-dimensional
> world. Why is that?
> 1) you have models of how things should be/are and so you can
> unambiguously say how their 2D projections will look like?
> 2) you see things in relationship to each other
> 3) You know how light shines and reflects of objects, and shadows and
> highlights tell you a lot.

> In fact the fact you have two eyes gives you almost no three
> dimensional advantage for small objects at distances more than twenty
> meters away. It is these other ways of perceiving three-D that let you
> see that.

Very good point. I believe the going fashion amongst cognitive
scientists is to say that the brain sees the world in "2 1/2"
dimensions -- an extraordinary approximation that allows us to function
almost flawlessly in a 3-D physical world. (For a fascinating but not
easy discussion of this topic, check out Steven Pinker's "How The Mind
Works".) The "almost" is also the explanation for many optical
illusions... which is why to my mind Murali is still kind of legal.
 
 
 

GE competition to catch cheats

Post by Gafoo » Fri, 21 Oct 2005 09:52:43

Quote:

> Can Murali or his lawyers sue the Shastra organizing board?

For what?
 
 
 

GE competition to catch cheats

Post by driz » Fri, 21 Oct 2005 10:18:25

Quote:
> Pataudi and Milburn had one eye and played a 3D ball reasonably well.

You are confusing two different things. Using two eyes to determine the
distance is one thing. Using two independent refrence frames to
calculate the angle between two lines is another thing. Our eyes, when
they focus on an object, form a triangle with the object as the 3rd
vertex. The internal angles of this triangle will be different based on
how far the object is. The angle variations are big with closer
objects, and diminish as the object goes farther. Therefore, in order
to differentiate the distance of two faraway objects, one has to have a
very good eye-sight (i.e., very good ability to corelate distance of
the object to the angle). We don't calculate or correlate all these
things in real life explicitly .. but our brain somehow performs all
this.

But the human brain is also very good at learning to do things a
different way. If you cover one eye and try playing the ball.. you'll
strugle initially. But you can adapt to measuring the distance and
speed of the ball based on other things. For instance, the position of
the ball relative to the pitch; or the change in apparant size of the
ball; or some reference point. Basically, it is possible for your brain
to learn these things in the absensce of one eye.

But this is not directly relevant to the topic here.

Quote:
>  You can close one eye and tell a lot about the three-dimensional
> world. Why is that?

The question here is not about 3-D or 2-D worlds. It is about whether
you can calculate the bending angle by looking side on only.

Quote:
> 1) you have models of how things should be/are and so you can
> unambiguously say how their 2D projections will look like?
>From 3-D to 2-D is straightforward. But the converse is not true. I can

give you an example.

Quote:
> 2) you see things in relationship to each other

To describe something on the 2-D plane, you need x and y coordinates.
Same way, to describe something in space you need x,y and z coordinate.
When u have a side-on camera, you only have x and y.

Quote:
> 3) You know how light shines and reflects of objects, and shadows and
> highlights tell you a lot.

Well .. how are you going to measure the shining from McGrath's skin?

Quote:

> In fact the fact you have two eyes gives you almost no three
> dimensional advantage for small objects at distances more than twenty
> meters away. It is these other ways of perceiving three-D that let you
> see that.

Agreed .. but not relevant to the question here.
 
 
 

GE competition to catch cheats

Post by dodo » Fri, 21 Oct 2005 10:32:50

Quote:

> > Pataudi and Milburn had one eye and played a 3D ball reasonably well.

> You are confusing two different things. Using two eyes to determine the
> distance is one thing. Using two independent refrence frames to
> calculate the angle between two lines is another thing. Our eyes, when
> they focus on an object, form a triangle with the object as the 3rd
> vertex. The internal angles of this triangle will be different based on
> how far the object is. The angle variations are big with closer
> objects, and diminish as the object goes farther. Therefore, in order
> to differentiate the distance of two faraway objects, one has to have a
> very good eye-sight (i.e., very good ability to corelate distance of
> the object to the angle). We don't calculate or correlate all these
> things in real life explicitly .. but our brain somehow performs all
> this.

> But the human brain is also very good at learning to do things a
> different way. If you cover one eye and try playing the ball.. you'll
> strugle initially. But you can adapt to measuring the distance and
> speed of the ball based on other things. For instance, the position of
> the ball relative to the pitch; or the change in apparant size of the
> ball; or some reference point. Basically, it is possible for your brain
> to learn these things in the absensce of one eye.

> But this is not directly relevant to the topic here.

> >  You can close one eye and tell a lot about the three-dimensional
> > world. Why is that?

> The question here is not about 3-D or 2-D worlds. It is about whether
> you can calculate the bending angle by looking side on only.

> > 1) you have models of how things should be/are and so you can
> > unambiguously say how their 2D projections will look like?

> >From 3-D to 2-D is straightforward. But the converse is not true. I can
> give you an example.

> > 2) you see things in relationship to each other

> To describe something on the 2-D plane, you need x and y coordinates.
> Same way, to describe something in space you need x,y and z coordinate.
> When u have a side-on camera, you only have x and y.

A first fact is that in camera projection  all straight lines project
as straight lines (except at image edges or with a fisheye camera)

So if you see that something is twisted side on, it probably is. The
question under the new ICC rules is how much it is.

If two lines meet at an angle they project at an angle that depends on
the view-point. However, if the viewpoint is side-on, and the camera is
far (e.g., aligned with the wickets on the boundary) the angle recorded
will be the true one.

However, this is not the forum for this discussion, and in any case
this is a matter of religion, and in any case the question has been
decided by the ICC.

- Show quoted text -

Quote:

> > 3) You know how light shines and reflects of objects, and shadows and
> > highlights tell you a lot.

> Well .. how are you going to measure the shining from McGrath's skin?

> > In fact the fact you have two eyes gives you almost no three
> > dimensional advantage for small objects at distances more than twenty
> > meters away. It is these other ways of perceiving three-D that let you
> > see that.

> Agreed .. but not relevant to the question here.

 
 
 

GE competition to catch cheats

Post by Colin Kynoc » Fri, 21 Oct 2005 10:54:23

On Thu, 20 Oct 2005 07:56:52 +1000, "Bounce It"

the following would improve the lives of those that read it:

Quote:



>>I agree with Colin ... just a lateral image is not sufficient to
>> conclude the bending.
>> Besides, the human skin color is not the same .. so the contraption
>> needs to address that also.

>> To me it seems like a BS undergrad project.

>So you understand more about this than the people working on the project.
>And your profession is???

I did high school mathematics, and that is all that is required to
know that you cannot tell if an elbow traveling through 3 dimensions
is straightening by taking a single 2 dimensional view of it.

If you don't believe this take two pencils and take them together at
an angle then imitate a bowling action.

From the side this can look like a straightening of the pencils, in
fact you can make those pencils that are at an angle look like they
are not at an angle if only one view point is taken into
consideration.  Then as you rotate them the angle will appear greater.

You do not need a degree in engineering or any similar field to see
that it is a novelty idea, which will tell you nothing.

Colin Kynoch

 
 
 

GE competition to catch cheats

Post by driz » Fri, 21 Oct 2005 13:44:43

Quote:

> A first fact is that in camera projection  all straight lines project
> as straight lines (except at image edges or with a fisheye camera)

Agreed. A perfectly straight arm will look straight from all angles.

Quote:
> So if you see that something is twisted side on, it probably is. The
> question under the new ICC rules is how much it is.

Agreed. Its the extent of bending that we are talking about. This will
change  depends on where you look from.

Quote:
> If two lines meet at an angle they project at an angle that depends on
> the view-point. However, if the viewpoint is side-on, and the camera is
> far (e.g., aligned with the wickets on the boundary) the angle recorded
> will be the true one.

This is where you are wrong. Your claim is correct ONLY IF the pivot is
also as far as the camera. But the pivot point is the shoulder here.

If you are mathematically inclined, work it out:
You have the point P(10,10,10) in the space. Connect the point P to
O(0,0,0).
First calculate the angle between the OP and the X-Y plane. Call it A1.
Now look from the x-direction (say, from (1000,0,0))and project OP on
the Y-Z plane. Calculate the angle between the projected OP and the X-Y
plane. Call it A2.

You'll see about 10 degree difference between A1 and A2.

This is a simple example to show the actual angle (A1) and apparant
angle from one side(A2) can be different. For a bowling action you have
two connected lines in the space (shoulder to elbow, and elbow to
wrist, let alone wrist to the ball). All three points (2 ends and the
joint) are moving. It is not a simple calculation dude. Single side-on
camera is a frikkin gross approximation!! You'll actually need at least
3 cameras from 3 orthogonal points. Or more than 3 from different
angles.

Have you ever wondered why the hawk-eye uses 6 cameras as opposed just
1 ? Why couldn't they take 10 pictures from behind the bowler and
extrapolate it ? Think ....