def inference(para_file): outdir = parameters.get_directory(para_file, 'inf_output_dir') # don't remove it automatically # if os.path.isdir(outdir): # io_function.delete_file_or_dir(outdir) # the script will check whether each image has been predicted command_string = os.path.join(eo_dir, 'workflow', 'parallel_prediction.py') + ' ' + para_file basic.os_system_exit_code(command_string)
def set_pythonpath(para_file): network_ini = parameters.get_string_parameters(para_file, 'network_setting_ini') mmseg_repo_dir = parameters.get_directory(network_ini, 'mmseg_repo_dir') mmseg_code_dir = osp.join(mmseg_repo_dir, 'mmseg') if os.path.isdir(mmseg_code_dir) is False: raise ValueError('%s does not exist' % mmseg_code_dir) # set PYTHONPATH to use my modified version of mmseg if os.getenv('PYTHONPATH'): os.environ['PYTHONPATH'] = os.getenv( 'PYTHONPATH') + ':' + mmseg_code_dir else: os.environ['PYTHONPATH'] = mmseg_code_dir print('\nPYTHONPATH is: ', os.getenv('PYTHONPATH'))
def mmseg_train_main(para_file, gpu_num): print(datetime.now(), "train MMSegmentation") SECONDS = time.time() if os.path.isfile(para_file) is False: raise IOError('File %s not exists in the current folder: %s' % (para_file, os.getcwd())) network_setting_ini = parameters.get_string_parameters( para_file, 'network_setting_ini') mmseg_repo_dir = parameters.get_directory(network_setting_ini, 'mmseg_repo_dir') mmseg_config_dir = osp.join(mmseg_repo_dir, 'configs') if os.path.isdir(mmseg_config_dir) is False: raise ValueError('%s does not exist' % mmseg_config_dir) base_config_file = parameters.get_string_parameters( network_setting_ini, 'base_config') base_config_file = os.path.join(mmseg_config_dir, base_config_file) if os.path.isfile(base_config_file) is False: raise IOError('%s does not exist' % base_config_file) global open_mmlab_python open_mmlab_python = parameters.get_file_path_parameters( network_setting_ini, 'open-mmlab-python') WORK_DIR = os.getcwd() expr_name = parameters.get_string_parameters(para_file, 'expr_name') # copy the base_config_file, then save to to a new one config_file = osp.join( WORK_DIR, osp.basename( io_function.get_name_by_adding_tail(base_config_file, expr_name))) if updated_config_file(WORK_DIR, expr_name, base_config_file, config_file, para_file, network_setting_ini, gpu_num) is False: raise ValueError('Getting the config file failed') train_evaluation_mmseg(WORK_DIR, mmseg_repo_dir, config_file, expr_name, para_file, network_setting_ini, gpu_num) duration = time.time() - SECONDS os.system( 'echo "$(date): time cost of training: %.2f seconds">>time_cost.txt' % duration)
def postProcess(para_file, inf_post_note, b_skip_getshp=False, test_id=None): # test_id is the related to training if os.path.isfile(para_file) is False: raise IOError('File %s not exists in current folder: %s' % (para_file, os.getcwd())) # the test string in 'exe.sh' test_note = inf_post_note WORK_DIR = os.getcwd() SECONDS = time.time() expr_name = parameters.get_string_parameters(para_file, 'expr_name') network_setting_ini = parameters.get_string_parameters( para_file, 'network_setting_ini') inf_dir = parameters.get_directory(para_file, 'inf_output_dir') if test_id is None: test_id = os.path.basename(WORK_DIR) + '_' + expr_name # get name of inference areas multi_inf_regions = parameters.get_string_list_parameters( para_file, 'inference_regions') # run post-processing parallel # max_parallel_postProc_task = 8 backup_dir = os.path.join(WORK_DIR, 'result_backup') io_function.mkdir(backup_dir) # loop each inference regions sub_tasks = [] same_area_time_inis = group_same_area_time_observations(multi_inf_regions) region_eva_reports = {} for key in same_area_time_inis.keys(): multi_observations = same_area_time_inis[key] area_name = parameters.get_string_parameters( multi_observations[0], 'area_name') # they have the same name and time area_time = parameters.get_string_parameters(multi_observations[0], 'area_time') merged_shp_list = [] map_raster_list_2d = [None] * len(multi_observations) for area_idx, area_ini in enumerate(multi_observations): area_remark = parameters.get_string_parameters( area_ini, 'area_remark') area_save_dir, shp_pre, _ = get_observation_save_dir_shp_pre( inf_dir, area_name, area_time, area_remark, test_id) # get image list inf_image_dir = parameters.get_directory(area_ini, 'inf_image_dir') # it is ok consider a file name as pattern and pass it the following functions to get file list inf_image_or_pattern = parameters.get_string_parameters( area_ini, 'inf_image_or_pattern') inf_img_list = io_function.get_file_list_by_pattern( inf_image_dir, inf_image_or_pattern) img_count = len(inf_img_list) if img_count < 1: raise ValueError( 'No image for inference, please check inf_image_dir and inf_image_or_pattern in %s' % area_ini) merged_shp = os.path.join(WORK_DIR, area_save_dir, shp_pre + '.shp') if b_skip_getshp: pass else: # post image one by one result_shp_list = [] map_raster_list = [] for img_idx, img_path in enumerate(inf_img_list): out_shp, out_raster = inf_results_to_shapefile( WORK_DIR, img_idx, area_save_dir, test_id) if out_shp is None or out_raster is None: continue result_shp_list.append(os.path.join(WORK_DIR, out_shp)) map_raster_list.append(out_raster) # merge shapefiles if merge_shape_files(result_shp_list, merged_shp) is False: continue map_raster_list_2d[area_idx] = map_raster_list merged_shp_list.append(merged_shp) if b_skip_getshp is False: # add occurrence to each polygons get_occurence_for_multi_observation(merged_shp_list) for area_idx, area_ini in enumerate(multi_observations): area_remark = parameters.get_string_parameters( area_ini, 'area_remark') area_save_dir, shp_pre, area_remark_time = get_observation_save_dir_shp_pre( inf_dir, area_name, area_time, area_remark, test_id) merged_shp = os.path.join(WORK_DIR, area_save_dir, shp_pre + '.shp') if os.path.isfile(merged_shp) is False: print('Warning, %s not exist, skip' % merged_shp) continue # add attributes to shapefile # add_attributes_script = os.path.join(code_dir,'datasets', 'get_polygon_attributes.py') shp_attributes = os.path.join(WORK_DIR, area_save_dir, shp_pre + '_post_NOrm.shp') # add_polygon_attributes(add_attributes_script,merged_shp, shp_attributes, para_file, area_ini ) add_polygon_attributes(merged_shp, shp_attributes, para_file, area_ini) # remove polygons # rm_polygon_script = os.path.join(code_dir,'datasets', 'remove_mappedPolygons.py') shp_post = os.path.join(WORK_DIR, area_save_dir, shp_pre + '_post.shp') # remove_polygons(rm_polygon_script,shp_attributes, shp_post, para_file) remove_polygons_main(shp_attributes, shp_post, para_file) # evaluate the mapping results # eval_shp_script = os.path.join(code_dir,'datasets', 'evaluation_result.py') out_report = os.path.join(WORK_DIR, area_save_dir, shp_pre + '_evaluation_report.txt') # evaluation_polygons(eval_shp_script, shp_post, para_file, area_ini,out_report) evaluation_polygons(shp_post, para_file, area_ini, out_report) ##### copy and backup files ###### # copy files to result_backup if len(test_note) > 0: backup_dir_area = os.path.join( backup_dir, area_name + '_' + area_remark_time + '_' + test_id + '_' + test_note) else: backup_dir_area = os.path.join( backup_dir, area_name + '_' + area_remark_time + '_' + test_id) io_function.mkdir(backup_dir_area) if len(test_note) > 0: bak_merged_shp = os.path.join( backup_dir_area, '_'.join([shp_pre, test_note]) + '.shp') bak_post_shp = os.path.join( backup_dir_area, '_'.join([shp_pre, 'post', test_note]) + '.shp') bak_eva_report = os.path.join( backup_dir_area, '_'.join([shp_pre, 'eva_report', test_note]) + '.txt') bak_area_ini = os.path.join( backup_dir_area, '_'.join([shp_pre, 'region', test_note]) + '.ini') else: bak_merged_shp = os.path.join(backup_dir_area, '_'.join([shp_pre]) + '.shp') bak_post_shp = os.path.join( backup_dir_area, '_'.join([shp_pre, 'post']) + '.shp') bak_eva_report = os.path.join( backup_dir_area, '_'.join([shp_pre, 'eva_report']) + '.txt') bak_area_ini = os.path.join( backup_dir_area, '_'.join([shp_pre, 'region']) + '.ini') io_function.copy_shape_file(merged_shp, bak_merged_shp) io_function.copy_shape_file(shp_post, bak_post_shp) if os.path.isfile(out_report): io_function.copy_file_to_dst(out_report, bak_eva_report, overwrite=True) io_function.copy_file_to_dst(area_ini, bak_area_ini, overwrite=True) # copy map raster b_backup_map_raster = parameters.get_bool_parameters_None_if_absence( area_ini, 'b_backup_map_raster') if b_backup_map_raster is True: if map_raster_list_2d[area_idx] is not None: for map_tif in map_raster_list_2d[area_idx]: bak_map_tif = os.path.join(backup_dir_area, os.path.basename(map_tif)) io_function.copy_file_to_dst(map_tif, bak_map_tif, overwrite=True) region_eva_reports[shp_pre] = bak_eva_report if len(test_note) > 0: bak_para_ini = os.path.join( backup_dir, '_'.join([test_id, 'para', test_note]) + '.ini') bak_network_ini = os.path.join( backup_dir, '_'.join([test_id, 'network', test_note]) + '.ini') bak_time_cost = os.path.join( backup_dir, '_'.join([test_id, 'time_cost', test_note]) + '.txt') else: bak_para_ini = os.path.join(backup_dir, '_'.join([test_id, 'para']) + '.ini') bak_network_ini = os.path.join(backup_dir, '_'.join([test_id, 'network']) + '.ini') bak_time_cost = os.path.join(backup_dir, '_'.join([test_id, 'time_cost']) + '.txt') io_function.copy_file_to_dst(para_file, bak_para_ini) io_function.copy_file_to_dst(network_setting_ini, bak_network_ini) if os.path.isfile('time_cost.txt'): io_function.copy_file_to_dst('time_cost.txt', bak_time_cost) # output the evaluation report to screen for key in region_eva_reports.keys(): report = region_eva_reports[key] if os.path.isfile(report) is False: continue print('evaluation report for %s:' % key) os.system('head -n 7 %s' % report) # output evaluation report to table if len(test_note) > 0: out_table = os.path.join( backup_dir, '_'.join([test_id, 'accuracy_table', test_note]) + '.xlsx') else: out_table = os.path.join( backup_dir, '_'.join([test_id, 'accuracy_table']) + '.xlsx') eva_reports = [ region_eva_reports[key] for key in region_eva_reports if os.path.isfile(region_eva_reports[key]) ] eva_report_to_tables.eva_reports_to_table(eva_reports, out_table) duration = time.time() - SECONDS os.system( 'echo "$(date): time cost of post-procesing: %.2f seconds">>time_cost.txt' % duration)
def main(options, args): print( "%s : prediction using the trained model (run parallel if use multiple GPUs) " % os.path.basename(sys.argv[0])) machine_name = os.uname()[1] start_time = datetime.datetime.now() para_file = args[0] if os.path.isfile(para_file) is False: raise IOError('File %s not exists in current folder: %s' % (para_file, os.getcwd())) basic.setlogfile('parallel_predict_Log.txt') deeplab_inf_script = os.path.join(code_dir, 'deeplabBased', 'deeplab_inference.py') network_setting_ini = parameters.get_string_parameters( para_file, 'network_setting_ini') global tf1x_python tf1x_python = parameters.get_file_path_parameters(network_setting_ini, 'tf1x_python') trained_model = options.trained_model outdir = parameters.get_directory(para_file, 'inf_output_dir') # remove previous results (let user remove this folder manually or in exe.sh folder) io_function.mkdir(outdir) # get name of inference areas multi_inf_regions = parameters.get_string_list_parameters( para_file, 'inference_regions') # max_parallel_inf_task = parameters.get_digit_parameters(para_file,'max_parallel_inf_task','int') b_use_multiGPUs = parameters.get_bool_parameters(para_file, 'b_use_multiGPUs') # loop each inference regions sub_tasks = [] for area_idx, area_ini in enumerate(multi_inf_regions): area_name = parameters.get_string_parameters(area_ini, 'area_name') area_remark = parameters.get_string_parameters(area_ini, 'area_remark') area_time = parameters.get_string_parameters(area_ini, 'area_time') inf_image_dir = parameters.get_directory(area_ini, 'inf_image_dir') # it is ok consider a file name as pattern and pass it the following functions to get file list inf_image_or_pattern = parameters.get_string_parameters( area_ini, 'inf_image_or_pattern') inf_img_list = io_function.get_file_list_by_pattern( inf_image_dir, inf_image_or_pattern) img_count = len(inf_img_list) if img_count < 1: raise ValueError( 'No image for inference, please check inf_image_dir and inf_image_or_pattern in %s' % area_ini) area_save_dir = os.path.join( outdir, area_name + '_' + area_remark + '_' + area_time) io_function.mkdir(area_save_dir) # parallel inference images for this area CUDA_VISIBLE_DEVICES = [] if 'CUDA_VISIBLE_DEVICES' in os.environ.keys(): CUDA_VISIBLE_DEVICES = [ int(item.strip()) for item in os.environ['CUDA_VISIBLE_DEVICES'].split(',') ] idx = 0 while idx < img_count: if b_use_multiGPUs: # get available GPUs # https://github.com/anderskm/gputil deviceIDs = GPUtil.getAvailable(order='first', limit=100, maxLoad=0.5, maxMemory=0.5, includeNan=False, excludeID=[], excludeUUID=[]) # only use the one in CUDA_VISIBLE_DEVICES if len(CUDA_VISIBLE_DEVICES) > 0: deviceIDs = [ item for item in deviceIDs if item in CUDA_VISIBLE_DEVICES ] basic.outputlogMessage('on ' + machine_name + ', available GPUs:' + str(deviceIDs) + ', among visible ones:' + str(CUDA_VISIBLE_DEVICES)) else: basic.outputlogMessage('on ' + machine_name + ', available GPUs:' + str(deviceIDs)) if len(deviceIDs) < 1: time.sleep( 60 ) # wait one minute, then check the available GPUs again continue # set only the first available visible gpuid = deviceIDs[0] basic.outputlogMessage( '%d: predict image %s on GPU %d of %s' % (idx, inf_img_list[idx], gpuid, machine_name)) else: gpuid = None basic.outputlogMessage('%d: predict image %s on %s' % (idx, inf_img_list[idx], machine_name)) # run inference img_save_dir = os.path.join(area_save_dir, 'I%d' % idx) inf_list_file = os.path.join(area_save_dir, '%d.txt' % idx) # if it already exist, then skip if os.path.isdir(img_save_dir) and is_file_exist_in_folder( img_save_dir): basic.outputlogMessage( 'folder of %dth image (%s) already exist, ' 'it has been predicted or is being predicted' % (idx, inf_img_list[idx])) idx += 1 continue with open(inf_list_file, 'w') as inf_obj: inf_obj.writelines(inf_img_list[idx] + '\n') sub_process = Process(target=predict_one_image_deeplab, args=(deeplab_inf_script, para_file, network_setting_ini, img_save_dir, inf_list_file, gpuid, trained_model)) sub_process.start() sub_tasks.append(sub_process) if b_use_multiGPUs is False: # wait until previous one finished while sub_process.is_alive(): time.sleep(5) idx += 1 # wait until predicted image patches exist or exceed 20 minutes time0 = time.time() elapsed_time = time.time() - time0 while elapsed_time < 20 * 60: elapsed_time = time.time() - time0 file_exist = is_file_exist_in_folder(img_save_dir) if file_exist is True or sub_process.is_alive() is False: break else: time.sleep(5) if sub_process.exitcode is not None and sub_process.exitcode != 0: sys.exit(1) # if 'chpc' in machine_name: # time.sleep(60) # wait 60 second on ITSC services # else: # time.sleep(10) # check all the tasks already finished while b_all_task_finish(sub_tasks) is False: basic.outputlogMessage('wait all tasks to finish') time.sleep(60) end_time = datetime.datetime.now() diff_time = end_time - start_time out_str = "%s: time cost of total parallel inference on %s: %d seconds" % ( str(end_time), machine_name, diff_time.seconds) basic.outputlogMessage(out_str) with open("time_cost.txt", 'a') as t_obj: t_obj.writelines(out_str + '\n')
def mmseg_parallel_predict_main(para_file, trained_model): print( "MMSegmetation prediction using the trained model (run parallel if use multiple GPUs)" ) machine_name = os.uname()[1] start_time = datetime.now() if os.path.isfile(para_file) is False: raise IOError('File %s not exists in current folder: %s' % (para_file, os.getcwd())) expr_name = parameters.get_string_parameters(para_file, 'expr_name') # network_ini = parameters.get_string_parameters(para_file, 'network_setting_ini') # mmseg_repo_dir = parameters.get_directory(network_ini, 'mmseg_repo_dir') # mmseg_code_dir = osp.join(mmseg_repo_dir,'mmseg') # if os.path.isdir(mmseg_code_dir) is False: # raise ValueError('%s does not exist' % mmseg_code_dir) # # set PYTHONPATH to use my modified version of mmseg # if os.getenv('PYTHONPATH'): # os.environ['PYTHONPATH'] = os.getenv('PYTHONPATH') + ':' + mmseg_code_dir # else: # os.environ['PYTHONPATH'] = mmseg_code_dir # print('\nPYTHONPATH is: ',os.getenv('PYTHONPATH')) if trained_model is None: trained_model = os.path.join(expr_name, 'latest.pth') outdir = parameters.get_directory(para_file, 'inf_output_dir') # remove previous results (let user remove this folder manually or in exe.sh folder) io_function.mkdir(outdir) # get name of inference areas multi_inf_regions = parameters.get_string_list_parameters( para_file, 'inference_regions') b_use_multiGPUs = parameters.get_bool_parameters(para_file, 'b_use_multiGPUs') # loop each inference regions sub_tasks = [] for area_idx, area_ini in enumerate(multi_inf_regions): area_name = parameters.get_string_parameters(area_ini, 'area_name') area_remark = parameters.get_string_parameters(area_ini, 'area_remark') area_time = parameters.get_string_parameters(area_ini, 'area_time') inf_image_dir = parameters.get_directory(area_ini, 'inf_image_dir') # it is ok consider a file name as pattern and pass it the following functions to get file list inf_image_or_pattern = parameters.get_string_parameters( area_ini, 'inf_image_or_pattern') inf_img_list = io_function.get_file_list_by_pattern( inf_image_dir, inf_image_or_pattern) img_count = len(inf_img_list) if img_count < 1: raise ValueError( 'No image for inference, please check inf_image_dir and inf_image_or_pattern in %s' % area_ini) area_save_dir = os.path.join( outdir, area_name + '_' + area_remark + '_' + area_time) io_function.mkdir(area_save_dir) # parallel inference images for this area CUDA_VISIBLE_DEVICES = [] if 'CUDA_VISIBLE_DEVICES' in os.environ.keys(): CUDA_VISIBLE_DEVICES = [ int(item.strip()) for item in os.environ['CUDA_VISIBLE_DEVICES'].split(',') ] idx = 0 while idx < img_count: if b_use_multiGPUs: # get available GPUs # https://github.com/anderskm/gputil # memory: orders the available GPU device ids by ascending memory usage deviceIDs = GPUtil.getAvailable(order='memory', limit=100, maxLoad=0.5, maxMemory=0.5, includeNan=False, excludeID=[], excludeUUID=[]) # only use the one in CUDA_VISIBLE_DEVICES if len(CUDA_VISIBLE_DEVICES) > 0: deviceIDs = [ item for item in deviceIDs if item in CUDA_VISIBLE_DEVICES ] basic.outputlogMessage('on ' + machine_name + ', available GPUs:' + str(deviceIDs) + ', among visible ones:' + str(CUDA_VISIBLE_DEVICES)) else: basic.outputlogMessage('on ' + machine_name + ', available GPUs:' + str(deviceIDs)) if len(deviceIDs) < 1: time.sleep( 60 ) # wait 60 seconds (mmseg need longer time to load models) , then check the available GPUs again continue # set only the first available visible gpuid = deviceIDs[0] basic.outputlogMessage( '%d: predict image %s on GPU %d of %s' % (idx, inf_img_list[idx], gpuid, machine_name)) else: gpuid = None basic.outputlogMessage('%d: predict image %s on %s' % (idx, inf_img_list[idx], machine_name)) # run inference img_save_dir = os.path.join(area_save_dir, 'I%d' % idx) inf_list_file = os.path.join(area_save_dir, '%d.txt' % idx) done_indicator = '%s_done' % inf_list_file if os.path.isfile(done_indicator): basic.outputlogMessage('warning, %s exist, skip prediction' % done_indicator) idx += 1 continue # if it already exist, then skip if os.path.isdir(img_save_dir) and is_file_exist_in_folder( img_save_dir): basic.outputlogMessage( 'folder of %dth image (%s) already exist, ' 'it has been predicted or is being predicted' % (idx, inf_img_list[idx])) idx += 1 continue with open(inf_list_file, 'w') as inf_obj: inf_obj.writelines(inf_img_list[idx] + '\n') sub_process = Process(target=predict_one_image_mmseg, args=(para_file, inf_img_list[idx], img_save_dir, inf_list_file, gpuid, trained_model)) sub_process.start() sub_tasks.append(sub_process) if b_use_multiGPUs is False: # wait until previous one finished while sub_process.is_alive(): time.sleep(1) idx += 1 # wait until predicted image patches exist or exceed 20 minutes time0 = time.time() elapsed_time = time.time() - time0 while elapsed_time < 20 * 60: elapsed_time = time.time() - time0 file_exist = os.path.isdir( img_save_dir) and is_file_exist_in_folder(img_save_dir) if file_exist is True or sub_process.is_alive() is False: break else: time.sleep(1) if sub_process.exitcode is not None and sub_process.exitcode != 0: sys.exit(1) basic.close_remove_completed_process(sub_tasks) # if 'chpc' in machine_name: # time.sleep(60) # wait 60 second on ITSC services # else: # time.sleep(10) # check all the tasks already finished wait_all_finish = 0 while basic.b_all_process_finish(sub_tasks) is False: if wait_all_finish % 100 == 0: basic.outputlogMessage('wait all tasks to finish') time.sleep(1) wait_all_finish += 1 basic.close_remove_completed_process(sub_tasks) end_time = datetime.now() diff_time = end_time - start_time out_str = "%s: time cost of total parallel inference on %s: %d seconds" % ( str(end_time), machine_name, diff_time.seconds) basic.outputlogMessage(out_str) with open("time_cost.txt", 'a') as t_obj: t_obj.writelines(out_str + '\n')
def image_translate_train_generate_main(para_file, gpu_num): ''' apply GAN to translate image from source domain to target domain existing sub-images (with sub-labels), these are image in source domain depend images for inference but no training data, each image for inference can be considered as on target domain ''' print(datetime.now(), "image translation (train and generate) using GAN") if os.path.isfile(para_file) is False: raise IOError('File %s not exists in current folder: %s' % (para_file, os.getcwd())) gan_para_file = parameters.get_string_parameters_None_if_absence( para_file, 'regions_n_setting_image_translation_ini') if gan_para_file is None: print( 'regions_n_setting_image_translation_ini is not set, skip image translation using GAN' ) return None gan_para_file = os.path.abspath( gan_para_file ) # change to absolute path, because later, we change folder training_regions = parameters.get_string_list_parameters( para_file, 'training_regions') machine_name = os.uname()[1] SECONDS = time.time() # get regions (equal to or subset of inference regions) need apply image translation multi_gan_regions = parameters.get_string_list_parameters( gan_para_file, 'regions_need_image_translation') multi_gan_source_regions = parameters.get_string_list_parameters( gan_para_file, 'source_domain_regions') # check target domain if len(multi_gan_source_regions) != len(multi_gan_regions): raise ValueError( 'the number of source domain and target domain is different') if set(multi_gan_source_regions).issubset(training_regions) is False: raise ValueError( 'the source domain regions are not the subset of training regions') for area_idx, (area_gan_ini, area_src_ini) in enumerate( zip(multi_gan_regions, multi_gan_source_regions)): basic.outputlogMessage('%d: source and target area: %s vs %s' % (area_idx, area_src_ini, area_gan_ini)) gan_working_dir = parameters.get_string_parameters(gan_para_file, 'working_root') # gan_dir_pre_name = parameters.get_string_parameters(gan_para_file, 'gan_dir_pre_name') # use GAN model name as the gan_dir_pre_name gan_model = parameters.get_string_parameters(gan_para_file, 'gan_model') gan_dir_pre_name = gan_model # '_' + # loop each regions need image translation sub_tasks = [] for area_idx, (area_gan_ini, area_src_ini) in enumerate( zip(multi_gan_regions, multi_gan_source_regions)): area_ini = os.path.abspath(area_gan_ini) area_src_ini = os.path.abspath(area_src_ini) area_name = parameters.get_string_parameters(area_ini, 'area_name') area_remark = parameters.get_string_parameters(area_ini, 'area_remark') area_time = parameters.get_string_parameters(area_ini, 'area_time') inf_image_dir = parameters.get_directory(area_ini, 'inf_image_dir') # it is ok consider a file name as pattern and pass it the following functions to get file list inf_image_or_pattern = parameters.get_string_parameters( area_ini, 'inf_image_or_pattern') inf_img_list = io_function.get_file_list_by_pattern( inf_image_dir, inf_image_or_pattern) img_count = len(inf_img_list) if img_count < 1: raise ValueError( 'No image for image translation, please check inf_image_dir and inf_image_or_pattern in %s' % area_ini) gan_project_save_dir = get_gan_project_save_dir( gan_working_dir, gan_dir_pre_name, area_name, area_remark, area_time, area_src_ini) if os.path.isdir(gan_project_save_dir): if generate_image_exists(gan_project_save_dir) is True: basic.outputlogMessage( 'generated new images (generate.txt_done) exist for %s exist, skip' % gan_project_save_dir) continue else: io_function.mkdir(gan_project_save_dir) # parallel run image translation for this area CUDA_VISIBLE_DEVICES = [] if 'CUDA_VISIBLE_DEVICES' in os.environ.keys(): CUDA_VISIBLE_DEVICES = [ int(item.strip()) for item in os.environ['CUDA_VISIBLE_DEVICES'].split(',') ] # get an valid GPU gpuids = [] while len(gpuids) < 1: # get available GPUs # https://github.com/anderskm/gputil deviceIDs = GPUtil.getAvailable(order='first', limit=100, maxLoad=0.5, maxMemory=0.5, includeNan=False, excludeID=[], excludeUUID=[]) # only use the one in CUDA_VISIBLE_DEVICES if len(CUDA_VISIBLE_DEVICES) > 0: deviceIDs = [ item for item in deviceIDs if item in CUDA_VISIBLE_DEVICES ] basic.outputlogMessage('on ' + machine_name + ', available GPUs:' + str(deviceIDs) + ', among visible ones:' + str(CUDA_VISIBLE_DEVICES)) else: basic.outputlogMessage('on ' + machine_name + ', available GPUs:' + str(deviceIDs)) if len(deviceIDs) < 1: print(datetime.now(), 'No available GPUs, will check again in 60 seconds') time.sleep( 60) # wait one minute, then check the available GPUs again continue # set only the first available visible gpuids.append(deviceIDs[0]) basic.outputlogMessage( '%d:image translation for %s on GPU %s of %s' % (area_idx, area_ini, str(gpuids), machine_name)) # run image translation # pytorch consider first GPUs in CUDA_VISIBLE_DEVICES as zero, so need to re-index gpu ids if len(CUDA_VISIBLE_DEVICES) > 0: gpuids = [CUDA_VISIBLE_DEVICES.index(id) for id in gpuids] sub_process = Process(target=image_translate_train_generate_one_domain, args=(gan_project_save_dir, gan_para_file, area_src_ini, area_ini, gpuids, inf_img_list)) sub_process.start() sub_tasks.append(sub_process) # wait until image translation has started or exceed 20 minutes time0 = time.time() elapsed_time = time.time() - time0 while elapsed_time < 20 * 60: elapsed_time = time.time() - time0 if CUT_gan_is_ready_to_train( gan_project_save_dir) is True or sub_process.is_alive( ) is False: break else: time.sleep(5) time.sleep( 10 ) # wait, allowing time for the GAN process to start, and run into problem if sub_process.exitcode is not None and sub_process.exitcode != 0: sys.exit(1) basic.close_remove_completed_process(sub_tasks) # check all the tasks already finished while basic.b_all_process_finish(sub_tasks) is False: basic.outputlogMessage('wait all tasks to finish') time.sleep(60) basic.check_exitcode_of_process(sub_tasks) basic.close_remove_completed_process(sub_tasks) save_image_dir = parameters.get_string_parameters(para_file, 'input_train_dir') save_label_dir = parameters.get_string_parameters(para_file, 'input_label_dir') merge_subImages_from_gan(multi_gan_source_regions, multi_gan_regions, gan_working_dir, gan_dir_pre_name, save_image_dir, save_label_dir) duration = time.time() - SECONDS os.system( 'echo "$(date): time cost of translating sub images to target domains: %.2f seconds">>time_cost.txt' % duration)
def image_translate_train_generate_one_domain(gan_working_dir, gan_para_file, area_src_ini, area_gan_ini, gpu_ids, domainB_imgList): current_dir = os.getcwd() # get orignal sub-images _, _, area_ini_sub_images_labels_dict = original_sub_images_labels_list_before_gan( ) sub_img_label_txt = os.path.join(current_dir, area_ini_sub_images_labels_dict) if os.path.isfile(area_ini_sub_images_labels_dict) is False: raise IOError( '%s not in the current folder, please get subImages first' % sub_img_label_txt) # prepare image list of domain A # what if the size of some images are not fit with CUT input? domain_A_images = [] # domain_A_labels = [] # with open(sub_img_label_txt) as txt_obj: # line_list = [name.strip() for name in txt_obj.readlines()] # for line in line_list: # sub_image, sub_label = line.split(':') # domain_A_images.append(os.path.join(current_dir,sub_image)) # # domain_A_labels.append(os.path.join(current_dir,sub_label)) area_ini_sub_images_labels = io_function.read_dict_from_txt_json( 'area_ini_sub_images_labels.txt') for line in area_ini_sub_images_labels[os.path.basename(area_src_ini)]: sub_image, sub_label = line.split(':') domain_A_images.append(os.path.join(current_dir, sub_image)) # domain_A_labels.append(os.path.join(current_dir,sub_label)) os.chdir(gan_working_dir) io_function.save_list_to_txt('image_A_list.txt', domain_A_images) # read target images, that will consider as target domains # what if there are too many images in domain B? io_function.save_list_to_txt('image_B_list.txt', domainB_imgList) gan_python = parameters.get_file_path_parameters(gan_para_file, 'python') cut_dir = parameters.get_directory(gan_para_file, 'gan_script_dir') train_script = os.path.join(cut_dir, 'train.py') generate_script = os.path.join(cut_dir, 'generate_image.py') # training of CUT if train_CUT_gan(gan_python, train_script, gan_para_file, gpu_ids) is False: os.chdir(current_dir) return False # genenerate image using CUT, convert images in domain A to domain B save_tran_img_folder = 'subImages_translate' if generate_image_CUT(gan_python, generate_script, gan_para_file, gpu_ids, domain_A_images, save_tran_img_folder) is False: os.chdir(current_dir) return False # change working directory back os.chdir(current_dir) pass