First steps towards bundle adjustment


The white points are triangulated from the correspondences in the first two video frames. The red shows the two camera positions. Green are from adding data from the third image. The third camera position is farther ahead because 3 frames are skipped in the video sequence.

This was done with three rectified jpeg images. When added to the Ladybug program, it will find correspondences from the raw images (higher resolution) and then project those points to the rectified position. This uses the camera intrinsics produced through the Ladybug API.

Python Enum continued

I needed to convert an enum to python so that I could decode some responses from C code. Here is an example of quickly converting a C enum to python.
The C code I am working with has several enum structures, one is a long list of error types:

enum LadybugError
{
   LADYBUG_OK,
   LADYBUG_FAILED,
   LADYBUG_INVALID_ARGUMENT,
   ...
   LADYBUG_INVALID_FRAMERATE,
   LADYBUG_INVALID_OFFSCREEN_BUFFER_SIZE,
   LADYBUG_INVALID_JPEG_IMAGE_STRUCTURE,

   LADYBUG_NUM_LADYBUG_ERRORS,

   LADYBUG_ERROR_FORCE_QUADLET = 0x7FFFFFFF,
}

There are about 70 items in this list.
Here is the conversion using namedtuple:

LadybugError = namedtuple('Enum', '''
   LADYBUG_OK,
   LADYBUG_FAILED,
   LADYBUG_INVALID_ARGUMENT,
   ...
   LADYBUG_INVALID_FRAMERATE,
   LADYBUG_INVALID_OFFSCREEN_BUFFER_SIZE,
   LADYBUG_INVALID_JPEG_IMAGE_STRUCTURE,

   LADYBUG_NUM_LADYBUG_ERRORS''')
LadybugError = LadybugError._make(range(len(LadybugError._fields)))

What is nice about namedtuple is that the keywords can be given as one long string. It parses the string for you by spaces and commas. This saves me the trouble adding quotes around each item, just add quotes around the whole list. I overwrite the original ‘LadybugError’ string with the enumeration because it is no longer needed.
When a program returns an integer representing an error, it can be passed to _fields[] to get the error string.

err = libc.ladybugGetStreamHeader( pContext, streamHeaderInfo, None )
print err, LadybugError._fields[err]

The output from this is:

20 LADYBUG_NOT_INITIALIZED

The value of err is 20 and ‘LADYBUG_NOT_INITIALIZED’ is the 21st item in the LadybugError enumeration.

Because I only need to interpret numerical feedback, I can drop the “_make(range” line and just take advantage of namedtuple’s builtin parsing:

LadybugError = namedtuple('Enum', '''
   LADYBUG_OK,
   LADYBUG_FAILED,
   LADYBUG_INVALID_ARGUMENT,
   ...
   LADYBUG_INVALID_FRAMERATE,
   LADYBUG_INVALID_OFFSCREEN_BUFFER_SIZE,
   LADYBUG_INVALID_JPEG_IMAGE_STRUCTURE,
   LADYBUG_NUM_LADYBUG_ERRORS''')

This helps me easily find out what an error of 12 means:

print error, LadybugError._fields[error]

prints

12 LADYBUG_COULD_NOT_OPEN_FILE

This works smoothly if the enumeration starts with zero. Otherwise, getting the name through _fields[i] will not be correct. Also, remember that exceptions can be added using ._replace(field=new_value). Here’s an example:

LadybugBusSpeed = LadybugBusSpeed._replace(LADYBUG_SPEED_UNKNOWN=-1)
err = -1
print LadybugBusSpeed._fields[LadybugBusSpeed.index(err)]