June 5, 2003

Love the availability, question the usability of open source apps

- by Lee Schlesinger -
Your task, Mr. Schlesinger, should you choose to accept it, is to scan the name of your high school classmates from an old newspaper and turn them into text for a class Web site. You may use either open source or Windows software, or a combination thereof. Good luck.

Well thanks, because I need it. I still haven't succeeded, but I have learned a few lessons.

Lesson 1: Windows works, but not perfectly. I first set up my Epson Perfection 1650 scanner on my Windows computer, because the included software CD came with drivers for Windows but not Linux. While the scanner was recognized just fine by Windows XP, the installation of the TWAIN driver necessary for scanning hung more than once, requiring more time and effort than expected. Once I completed the installation successfully, the Epson SmartPanel application neatly presented all the possible scanning tasks and walked users through using them.

I was able to scan the old pages of newsprint just fine under Windows. However, the OCR application provided with the Epson scanner completely failed to recognize any text.

Lesson 2: There are open source applications for almost everything. I searched the Web and found two applications I thought could help. Kooka is a KDE-based scanner front end, while gocr is a GNU OCR program.

Lesson 3: Don't fear the compiler. This was probably the most useful lesson from the whole affair. I've been hesitant to compile applications from source, preferring for some reason to install only binaries. Following the instructions in each application's documentation showed me that compiling from source and installing the resulting applications is almost trivial. You just download the source archive and decompress it, both easily done through the graphical user interface. Then, open a console window, type ./configure, make, su, and make install, and there you are. You can use the command whereis to verify the install completed successfully.

Of course, it turned out I probably didn't need to learn that lesson. A little more reading showed me that Kooka is installed by default with KDE, and was probably on my Mandrake 9.1 system before I undertook the exercise. Clearly I'm still learning. I was also happy to discover that Mandrake's ScannerDrake tool recognized my USB scanner and installed the proper drivers for me.

Lesson 4: Linux applications are no more stable than apps under Windows. Kooka did a decent job of scanning my pages. Before I OCRed them, however, I had to rotate them. Every time I tried, Kooka "crashed and caused the signal 11 (SIGSEGV)."

Okay, there's more than one way to skin a cat. What if I rotated the images using the Gimp, then OCRed them? The Gimp did the job just fine. It was the second part of the task that stopped me. Kooka can't read an existing graphic file and pass it on to the OCR engine. Kooka developers, this would be a nice enhancement.

Still, I thought, I can turn to gocr on the command line. And sure enough, gocr processed the .pnm file Kooka had scanned and the Gimp had rotated. Unfortunately, what resulted was a meaningless jumble of characters.

Lesson 5: Sometimes the old-fashioned way is the only way. I guess I'll have to type in all 551 names by hand. Good thing we have four more years until the next reunion. Or -- wait! -- what about voice recognition ...


  • Linux
Click Here!