CSC 249/449 Computer Vision: Self Portrait Assignment


Weighting factor 2.

FIRST READ THE RESOURCES PAGE.

This assignment is to take some pictures and store them on the CS graduate system in a specific form and format. This will require you to get your account on the graduate system, and find at least one way of getting images into it, as well as learing a little bit about image formats, conversion, and scaling. When you get your account from Jim Roche (CSC 630), you should ask for card access to the vision and software labs.

Use the equipment in the vision lab (CSC 616), the software lab (CSC 727), or your own camera to take two images of yourself against a neutral background. The images should resemble standard ID photos - that is, frontal face images that fill a sizable portion of the picture. For purposes of standardization, make the images 360 columns by 480 rows. The program "display" in Redhat linux systems, (part of ImageMagick) or my program "image_calc" (read about it in the resources pages). or other tools can be used to trim and scale your pics and convert formats Don't scale the rows and columns differently. We may use these pictures later in a face identification projects, so don't use bizarre lighting or extreme facial expressions. One image should be color, the other monochrome, and you need to get them into UNCOMPRESSED TIFF format, 24-bit for the color image, 8-bit for the monochrome. You can obtain the monochrome image either using a monochrome camera, or by derivation from a color image, but the pictures should be different shots (though you don't have to go to any particular effort to make them different).

Once you are happy with your images, you need to put them in a directory that will serve as a central resource for this course. The directory is /u/nelson/pics/faces/csc_449_spring_2010. You should be able to write to it, but will be able to modify only files that you own. The image files should be constructed using your name following the model: first_last.mono.tiff and first_last.color.tiff (e.g., randal_nelson.color.tiff). No capital letters. Remember, you must get your images into UNCOMPRESSED TIFF. If you used your digital camera, that is probably not what the original format is.

Read the web pages under resources on imagery, cameras, and digitizers before you attempt to do any of this. There are color and monochrome cameras in the vision lab that you can use, also a couple of camcorders that you can probably use as well, though it may take some additional mucking around. There are BT848-based digitizers in the shellfish, whale-named and s* linux machines in the 6th-floor vision lab that take S-video and composite. The cards are accessible from the back of the computer cases, and each has 2 or 3 RCA composite input ports, and one S-video input port. If you know what these look like, that is enough to recognize the card. Otherwise get someone to show you. Basically, figuring out how to wire up and use all this is up to you. Some of the cards get color images only on the composite channel. This should be noted on a sticky note, but if you find you are getting black and white on the s-video channel, try the composite.

You can use your own, or some other equipment to obtain the images for this assignment if you want, but in that case you may miss out on some crucial learning that could be critical later in the course if there is an assignment for which your equipment won't work. Using a pre-existing digital image is not legit for this assignment; digitizing a hard-copy photograph is ok. Be careful with converting highly compressed formats (as are likely to be used on your cell phome camera). If the full res version doesn't lool clear and artifact-free, don't use it. The same caution applies if you have software on your machine that produces gif files, which are 8-bit color mapped pictures, or any other compressed format (e.g. most JPEG), unless it is lossless compression. (if you don't know what that is or if you have it, you can assume you don't)

If you are uneasy about providing actual pictures of yourself for any reason, you are welcome to use another person, realistic doll, or even a movie star as standin, assuming you can get multiple images of the sort needed, now and in the future.

The general intent of this exercise is to get you familiar with the use of digital cameras, dealing with digitizers, and use of at least one digital image format. A large part of this assignment is exercising your ability to obtain the information you need to do this from available sources. These include the resources pages on the course web site, your classmates (it's ok to ask for this assignment), more senior graduate students, man pages, "Readme" files, system library includes and source code, etc. The course instructor (that's me) is not fair game, except in the case of broken or missing equipment (or code), in which case you should let me know. Even in this case, you should try to be sure that the artifacts are actually broken or missing, and you did not forget to turn on a switch, plug something in, use the right channel, or include the right files.

Hand in a printout of your two pictures, and a short writeup decribing how you took the images (what equipment you used etc.)

Back to vision course main page