end = timer() mve_makescene_function_time = end - start start = timer() lib.mve_dense_reconstruction(submodel_path ,max_concurrency, True, file_path) end = timer() mve_dense_reconstruction_time = end - start start = timer() lib.mve_scene2pset_function(submodel_path, max_concurrency, True, file_path) end = timer() mve_mve_scene2pset_time = end - start start = timer() lib.mve_cleanmesh_function(submodel_path, max_concurrency) end = timer() mve_mve_cleanmesh_time = end - start start = timer()
def do_task(job_queue): while True: try: task = job_queue.get() # returns a map print("task: " + str(task)) if task['clientid'] == task['nodeid']: print('equals self node id') response = False if task['title'] == 'upload_image': #check if there is a is a submodel folder print('no need to upload for self') if task['title'] == 'extract_exif': #image path, opensfm_config, node_file_path print('extract exif in self') # file_path print(task['node_file_path']) image_path = os.path.join(task['node_file_path'], 'images') response = lib.sfm_extract_metadata_list_of_images(image_path, task['opensfm_config'], task['node_file_path'], task['imagelist']) print('response in self') print(response) elif task['title'] == 'detect_features': print('detect features in task queue in self') # needs images # current path, ref images, opensfm_config current_path = task['node_file_path'] ref_image = os.listdir(current_path + '/images') response = lib.sfm_detect_features(ref_image, current_path ,task['opensfm_config'], task['imagelist']) elif task['title'] == 'feature_matching': print('feature matching') # needs images # image path # list of images # candidate images # opensfm config response = lib.sfm_feature_matching(task['node_file_path'], task['ref_image'], task['cand_images'], task['opensfm_config']) elif task['title'] == 'feature_matching_pairs': print('in task here ' + str(task['pair'])) pair = task['pair'] #send the pair of feature files feature_0 = 'features/' + pair[0] + '.features.npz' feature_1 = 'features/' + pair[1] + '.features.npz' #response = task['client'].upload(task['filepath'], [feature_0, feature_1]) exif_0 = 'exif/' + pair[0] + '.exif' exif_1 = 'exif/' + pair[1] + '.exif' image_list = [pair[0], pair[1]] pairs_matches = lib.sfm_feature_matching_pairs(task['filepath'],image_list,image_list ,task['opensfm_config']) filename = pair[0]+'-'+pair[1] # save pair matches in a file # send it to client opensfm_interface.save_matches(task['filepath'], filename, pairs_matches) #filename = pair[0]+'-'+pair[1] matches = opensfm_interface.load_matches(task['filepath'],filename) #print('load matches') #print(matches) task['lock'].acquire() try: print('lock acquired') tup = (pair[0], pair[1]) task['results'].update(matches) print('updated matches') except Exception as e: print(e.message) pass task['lock'].release() print('lock release') elif task['title'] == 'sfm_create_tracks': # needs images # current path #submodel version print(task['cluster']) sending_match_pairs = {} print(task['results'].keys()) print('here') for key, value in task['results'].items(): #print(key) if key[0] in task['cluster'] and key[1] in task['cluster']: print('key:' + str(key)) sending_match_pairs[key] = value print(len(sending_match_pairs)) print(sending_match_pairs) print(os.path.join(task['filepath'], task['submodel_path'])) matches_filepath =os.path.join(task['filepath'], task['submodel_path']) # maybe save the results in a results file and send it submodel_path = task['submodel_path'] print(submodel_path) #opensfm_interface.save_matches(matches_filepath, submodel_matches, sending_match_pairs) #matches = opensfm_interface.load_matches(matches_filepath, submodel_matches) #print(matches) response = lib.sfm_create_tracks(submodel_path, task['cluster'],task['opensfm_config'], sending_match_pairs, True, task['filepath']) start = timer() # # need the exifs lib.sfm_opensfm_reconstruction(submodel_path, opensfm_config, True,task['filepath'] ) end = timer() sfm_opensfm_reconstruction_time = end - start start = timer() lib.sfm_undistort_image(submodel_path, opensfm_config, True, task['filepath']) end = timer() sfm_undistort_image_time = end - start start = timer() lib.sfm_export_visual_sfm(submodel_path, opensfm_config ,True, task['filepath']) end = timer() sfm_export_visualsfm_time = end - start start = timer() #test load udata = opensfm_interface.UndistortedDataSet(task['filepath'], task['opensfm_config'], 'undistorted') for image_name in task['cluster']: try: #print(image_name) ret = udata.load_raw_depthmap(image_name) except Exception as e: print(e.message) print('image name bad ' + str(image_name)) lib.sfm_compute_depthmaps(submodel_path, opensfm_config, True, task['filepath']) end = timer() sfm_compute_depthmaps_time = end - start # split merge here no problem start = timer() max_concurrency = 2 #need images # delete from makescene #lib.mve_makescene_function(submodel_path, max_concurrency, True, task['file_path']) end = timer() mve_makescene_function_time = end - start start = timer() #lib.mve_dense_reconstruction(submodel_path ,max_concurrency, True, task['file_path']) end = timer() mve_dense_reconstruction_time = end - start start = timer() #lib.mve_scene2pset_function(submodel_path, max_concurrency) end = timer() mve_mve_scene2pset_time = end - start start = timer() # lib.mve_cleanmesh_function(submodel_path, max_concurrency) end = timer() mve_mve_cleanmesh_time = end - start start = timer() # lib.odm_filterpoints_function(submodel_path, max_concurrency) end = timer() odm_filterpoint_time = end - start start = timer() # lib.odm_mesh_function(submodel_path, max_concurrency) end = timer() odm_mesh_time = end - start print('OpenSfm Reconstruction Total Time: ' + str(sfm_opensfm_reconstruction_time)) print('OpenSfm Undistort Image Total Time: ' + str(sfm_undistort_image_time)) print('OpenSfm Export Visual Sfm Total Time: ' + str(sfm_export_visualsfm_time)) print('OpenSfm Compute DepthMaps Sfm Total Time: ' + str(sfm_compute_depthmaps_time)) print('Mve Makescene Sfm Total Time: ' + str(mve_makescene_function_time)) print('Mve Dense Reconstruction Sfm Total Time: ' + str(mve_dense_reconstruction_time)) print('Mve Scene 2 Pset Sfm Total Time: ' + str(mve_mve_scene2pset_time)) print('Mve Cleanmesh Sfm Total Time: ' + str(mve_mve_cleanmesh_time)) print('ODM Filterpoints Total Time: ' + str(odm_filterpoint_time)) print('ODM Mesh Total Time: ' + str(odm_mesh_time)) elif task['title'] == 'opensfm_reconstruction': # needs create tracks . csv # return reconstruction.json response = lib.sfm_opensfm_reconstruction(task['node_file_path'], task['opensfm_config']) elif task['title'] == 'sfm_undistort_image': response = lib.sfm_undistort_image(task['node_file_path'], task['opensfm_config']) elif task['title'] == 'export_visual_sfm': #nvm file response = lib.sfm_export_visual_sfm(task['node_file_path'],task['opensfm_config']) elif task['title'] == 'compute_depthmaps': response = lib.sfm_compute_depthmaps(task['node_file_path']. task['opensfm_config']) elif task['title'] == 'dense_reconstruction': response = lib.mve_dense_reconstruction(task['node_file_path']) elif task['title'] == 'makescene_function': response = lib.mve_makescene_function(task['node_file_path'], task['max_concurreny']) elif task['title'] == 'scene2pset': response = lib.mve_scene2pset_function(task['node_file_path'], task['max_concurreny']) elif task['title'] == 'cleanmesh': response = lib.mve_cleanmesh_function(task['node_file_path'], task['max_concurreny']) elif task['title'] == 'filterpoints': response = lib.odm_filterpoints_function(task['node_file_path'], task['max_concurreny']) elif task['title'] == 'mesh': response = lib.odm_mesh_function(task['node_file_path'], task['max_concurreny']) elif task['title'] == 'texturing': response = lib.odm_texturing_function(task['node_file_path']) else: if task['title'] == 'upload_image': print('upload') response = task['client'].upload(task['filepath'], task['list']) print(response) elif task['title'] == 'extract_exif': print(str(task)) print('extract exif ') response = task['client'].sendTask('exif', task['nodeid'], '/exif') print(response) elif task['title'] == 'detect_features': print('detect features in task queue') # needs images response = task['client'].sendTask('detect_features', task['nodeid'], '/features') elif task['title'] == 'feature_matching': print('feature matching') # needs images respone = task['client'].sendTask('feature_matching', task['nodeid'], '/matches') elif task['title'] == 'sfm_create_tracks': #needs images print(task['cluster']) sending_match_pairs = {} print(task['results'].keys()) print('here') for key, value in task['results'].items(): #print(key) if key[0] in task['cluster'] and key[1] in task['cluster']: print('key:' + str(key)) sending_match_pairs[key] = value print(len(sending_match_pairs)) print(sending_match_pairs) print(os.path.join(task['filepath'], task['submodel_path'])) matches_filepath =os.path.join(task['filepath'], task['submodel_path']) # maybe save the results in a results file and send it submodel_matches = task['submodel_path'] print(submodel_matches) opensfm_interface.save_matches(matches_filepath, submodel_matches, sending_match_pairs) #matches = opensfm_interface.load_matches(matches_filepath, submodel_matches) print(matches) print(' upload matches ') feature_matches_response = task['client'].upload(matches_filepath, [submodel_matches+'_matches.pkl.gz'], submodel_file='matches') print('here after feature matches send') # send features # send matches feature_list = [] for image_name in task['cluster']: feature_0 = 'features/' + image_name + '.features.npz' feature_list.append(feature_0) print(' send features ') response = task['client'].upload(task['filepath'], feature_list, 'feature', task['submodel_path'] ) #send images response = task['client'].upload(task['filepath'], task['cluster'], 'images', task['submodel_path']) #send exif response = task['client'].upload(task['filepath'], task['cluster'], 'exif', task['submodel_path']) #send cameras.json response = task['client'].upload(task['filepath'], ['camera_models.json'], 'camera_models', task['submodel_path'] ) # respone = task['client'].sendTask('create_tracks', task['nodeid'], '/', submodel_path=task['submodel_path'], cluster_images = task['cluster']) elif task['title'] == 'opensfm_reconstruction': # needs create tracks . csv # return reconstruction.json respone = task['client'].sendTask('opensfm_reconstruction', task['nodeid'], '/') elif task['title'] == 'sfm_undistort_image': respone = task['client'].sendTask('undistort_image', task['nodeid'], '/undistorted') elif task['title'] == 'export_visual_sfm': #nvm file respone = task['client'].sendTask('export_visual_sfm', task['nodeid'], '/undistorted') elif task['title'] == 'compute_depthmaps': respone = task['client'].sendTask('compute_depthmaps', task['nodeid'], '/depthmaps') elif task['title'] == 'dense_reconstruction': respone = task['client'].sendTask('dense_reconstrution', task['nodeid'], '/mve') elif task['title'] == 'makescene_function': respone = task['client'].sendTask('makescene_function', task['nodeid'], '/mve') elif task['title'] == 'scene2pset': respone = task['client'].sendTask('scene2pset', task['nodeid'], '/mve') elif task['title'] == 'cleanmesh': respone = task['client'].sendTask('cleanmesh', task['nodeid'], '/mve') elif task['title'] == 'filterpoints': respone = task['client'].sendTask('odm_filterpoints', task['nodeid'], '/filterpoints') elif task['title'] == 'mesh': respone = task['client'].sendTask('odm_mesh', task['nodeid'], '/mesh') elif task['title'] == 'texturing': respone = task['client'].sendTask('odm_texturing', task['nodeid'], '/mvs') elif task['title'] == 'feature_matching_pairs': # send over the pair feature files and exif files print('in task here ' + str(task['pair'])) pair = task['pair'] #send the pair of feature files feature_0 = 'features/' + pair[0] + '.features.npz' feature_1 = 'features/' + pair[1] + '.features.npz' response = task['client'].upload(task['filepath'], [feature_0, feature_1]) exif_0 = 'exif/' + pair[0] + '.exif' exif_1 = 'exif/' + pair[1] + '.exif' response = task['client'].upload(task['filepath'], [exif_0, exif_1]) response = task['client'].sendTask('feature_match_pairs', task['nodeid'], '/matches', [pair[0], pair[1]]) filename = pair[0]+'-'+pair[1] matches = opensfm_interface.load_matches(task['filepath'],filename) #print('load matches') #print(matches) print('try lock') task['lock'].acquire() try: tup = (pair[0], pair[1]) task['results'].update(matches) print('lock acquired and reuslt done for send ') except Exception as e: print(e.message) pass task['lock'].release() print('lock release for send ') print('done') job_queue.task_done() except Exception as e: print(e.message) print(traceback.print_exc()) print() sys.exit(0) except KeyboardInterrupt: sys.exit(0)
# delete from makescene lib.mve_makescene_function(current_path, max_concurrency) end = timer() mve_makescene_function_time = end - start start = timer() lib.mve_dense_reconstruction(current_path, max_concurrency) end = timer() mve_dense_reconstruction_time = end - start start = timer() lib.mve_scene2pset_function(current_path, max_concurrency) end = timer() mve_mve_scene2pset_time = end - start start = timer() lib.mve_cleanmesh_function(current_path, max_concurrency) end = timer() mve_mve_cleanmesh_time = end - start start = timer() lib.odm_filterpoints_function(current_path, max_concurrency)