naptha
Add a way to set "Init Only" parameters (user_word_suffix, etc.)
I would like to use the command-line parameters "user_word_suffix", "load_freq_dawg", and "load_system_dawg". After sorting through a lot of documentation, and looking through a lot of code, I realized that these are "init only" parameters. In the TessBaseAPI code, they need to be passed to Init(), either as a set of keys/values or in a config file. Setting the parameters after initialization doesn't work because the traineddata files have already been read and the dictionaries formed.
Suggested fix: Add a config filename optional parameter (string) to worker.Initialize(...) that gets passed to api.Init(...).
Other fixes: Add a worker.SetInitParameters() function, just like worker.SetParameters(), that must be called before worker.Initialize, and pass those keys/values to api.Init(). Add a "initParams" optional parameter to worker.Initialize, which contains key/value pairs that get passed to api.Init
I'm suggesting the config file option because it feels like the least work to get the desired result.
I looked it up, and it seems like the init-only parameters are few in number and relatively fringe (notably disabling various dictionaries). However, I agree that it would be nice to have some way for advanced users to specify a config file (like is possible on desktop). I will look into whether this can be easily added in a future release.
This feature has been added to the dev/v4 branch, and will be released with version 4. If you would like to test before then, instructions are in #662.
To easily verify that these options are indeed being set, I am attaching a test image with significantly different results for the legacy model (oem: "0") depending on whether load_number_dawg is enabled.
Results with load_number_dawg: "1"
1823747 72460000
271.83 1223.00
3164675.10 1512895284
Results with load_number_dawg: "0"
18237.47 724600.00
271.83 1223.00
3164675.10 15128952.84