Run_image.py command drops out

[Comando1026]

Running a full build of the mapkurator pipeline on a wsl ubuntu 18.04, and attempted to run the crop, spotter etc using altered version of the command listed on the website (see below)

python run_img.py --map_kurator_system_dir /home/mapkurator/mapkurator-system/ --input_dir_path /mnt/c/Users/ethan/OneDrive/Desktop/PHD-OCR/Detection\ Model/Input --expt_name mapKurator_test --module_cropping --module_get_dimension --module_text_spotting --text_spotting_model_dir /home/mapkurator/mapkurator-spotter/spotter-v2/ --spotter_model spotter-v2 --spotter_config /home/mapkurator/mapkurator-spotter/spotter-v2/configs/pretrain/pretrain_en_synthtext_synthmap_tt_ic_mlt.yaml --spotter_expt_name test --module_img_geojson --output_folder /mnt/c/Users/ethan/OneDrive/Desktop/PHD-OCR/Detection\ Model/Output --gpu_id 0

command runs without throwing initial errors and produces the crop, spotter,stitch output folders in the correct paths along with a empty error .csv before dropping out without a clear reason. only deprecation warnings for pandas and easy_install are given in the command line. No data is produced

Input files attempted has been .tif, .jpg, .png and a .jp2 without any success

any help would be greatly appreciated!!

Additionally the initial code user a config in PALEJUN known as Rumsey but no folder or config file of those names can be found in the mapkurator-spotter setup/install

[zekun-li]

If you're using the mapKurator within the docker image, you might want to check if the input_dir_path accessible within the docker. You can do a ls command to list the content in that folder. If it is not accessible, you'll want to use -v to expose the folder when doing docker run to launch the image

[Comando1026]

Thanks for the reply, I am using mapKurator outside of the docker image/ running it locally as I want to utilise both the detection and recognition modules, upon checking the ls of the input_dir_path it does seem to be accessible by the program and it can detect the two images within the folder. (See image above)

Thanks for the assistance!

[zekun-li]

If nothing has been produced in the output crop folder, it means the system did not succeed in executing the first module. If you are able to edit the utils.py file, you can either add breakpoints or add pdb.set_trace() to check if the sample_map_df stores the correct image paths. Similarly, you can also use the same technique in the cropping and spotting module functions below in the same python file.

[Comando1026]

Good news, sample_map_df is populating as expected, and up to the cropping module, where it seems that the cropping python isnt called correctly or something is going on with it.

            run_crop_command = 'python crop_img.py --img_path '+img_path + ' --output_dir '+ cropping_output_dir
            pdb.set_trace()
            exe_ret = execute_command(run_crop_command, if_print_command)

pdb trace tests

(Pdb) p img_path
'/mnt/c/Users/ethan/OneDrive/Desktop/PHD-OCR/Detection Model/Input/Sydney 1877.jpeg'

(Pdb) p run_crop_command
'python crop_img.py --img_path /mnt/c/Users/ethan/OneDrive/Desktop/PHD-OCR/Detection Model/Input/Sydney 1877.jpeg --output_dir /mnt/c/Users/ethan/OneDrive/Desktop/PHD-OCR/Detection Model/Output/mapKurator_test/crop/'
(Pdb) c

> /home/mapkurator/mapkurator-system/utils.py(188)run_pipeline()
-> exe_ret = execute_command(run_crop_command, if_print_command)

(Pdb) p img_path
'/mnt/c/Users/ethan/OneDrive/Desktop/PHD-OCR/Detection Model/Input/Sydney 1877.jpg'
(Pdb) p run_crop_command
``
'python crop_img.py --img_path /mnt/c/Users/ethan/OneDrive/Desktop/PHD-OCR/Detection Model/Input/Sydney 1877.jpg --output_dir /mnt/c/Users/ethan/OneDrive/Desktop/PHD-OCR/Detection Model/Output/mapKurator_test/crop/'
(Pdb)

I have setup pdb tests on the crop_img.py file but none triggered

Found Error

'usage: crop_img.py [-h] [--img_path IMG_PATH] [--output_dir OUTPUT_DIR]\tcrop_img.py: error: unrecognized arguments: Model/Input/Sydney 1877.jpeg Model/Output/mapKurator_test/crop/\t'
(Pdb)

Appears the parsing of the command is not working as intended

Fix

changed run_crop_command to

run_crop_command = ["python", "crop_img.py", "--img_path", img_path, "--output_dir", cropping_output_dir]

Similar Changes Below:

run_spotting_command

run_spotting_command = ["python", "tools/inference.py", "--config-file", spotter_config,"--output_json", "--input", os.path.join(cropping_output_dir,map_name), "--output",map_spotting_output_dir]

run_stitch_command

run_stitch_command = ['python','stich_output.py', "--input_dir", stitch_input_dir, "--output_geojson", output_geojson]

Switching the models to the Rumsey.yml extracted from the docker image fixed the spotter and stitch issues, so the code now works for crop->stitch perfectly

Any help getting the post-ocr working would be amazing tho!!

[zekun-li]

Glad that it's working now! Looks like the space in filenames was causing the problem, and your modified code provides a work-around for that.

Post-OCR module requires some database set-ups and preprocessing, and the steps are here