api/synchronous/api_core/animal_detection_classification_api/runserver.py (3 lines):
	- line 41: # TODO classifier = TFClassifier(api_config.CLASSIFICATION_MODEL_PATHS, api_config.CLASSIFICATION_CLASS_NAMES)
	- line 85: print('runserver, post_detect_sync, user specified detection confidence: ', detection_confidence)  # TODO delete
	- line 257: # TODO 'classification mean inference time': str(''),


api/batch_processing/api_core/server_utils.py (3 lines):
	- line 22: # TODO log exception when we have more telemetry
	- line 31: # TODO check that the expiry date of input_container_sas is at least a month
	- line 41: # TODO - check based on access policy as well


api_flask_redis/api_core/animal_detection_api/api_frontend.py (2 lines):
	- line 151: # TODO: read from memory rather than using intermediate files
	- line 169: # TODO: convert to a blocking read and eliminate the sleep() statement in this loop


classification/efficientnet/utils.py (2 lines):
	- line 93: # TODO: modify the params names.
	- line 576: # TODO: add the petrained weights url map of 'efficientnet-l2'


classification/train_utils.py (2 lines):
	- line 56: TODO: potential speedup by avoiding PNG compression and PIL dependency
	- line 239: # TODO: consider weighting val and test set as well


taxonomy_mapping/species_by_dataset.py (2 lines):
	- line 261: for i_row, cell in enumerate(ws['D']):  # TODO hardcoded column number
	- line 267: # TODO hardcoded columns: change if # of examples or col_order changes


classification/train_classifier_tf.py (2 lines):
	- line 73: Possible TODO: oversample the imbalanced classes
	- line 345: # TODO: change weighted to False if oversampling minority classes


research/active_learning/labeling_tool/runapp.py (1 line):
	- line 73: args.strategy = 'confidence' ## TODO: hard coding this for now


research/active_learning/Database/initialize_target_db.py (1 line):
	- line 69: ## TODO: allow user to update the class list through the labeling tool UI as they see different species


research/airsim/get_gt_from_images.py (1 line):
	- line 65: s = img.shape; h = s[0]; w = s[1]  #TODO: check this isn't backwards


api/batch_processing/integration/eMammal/WPF-integration-app/eMammalIntegration.cs (1 line):
	- line 232: // TODO: confirm json file is reading in detections correctly


data_management/importers/snapshot_serengeti_lila.py (1 line):
	- line 123: # TODO: iterrows() is a terrible way to do this, but this is one of those days


api/batch_processing/api_core/server.py (1 line):
	- line 108: if model_version not in api_config.MD_VERSIONS_TO_REL_PATH:  # TODO use AppConfig to store model version info


api_flask_redis/api_core/animal_detection_api/api_backend.py (1 line):
	- line 35: # TODO: convert to a blocking read and eliminate the sleep() statement in this loop


research/active_learning/labeling_tool/static/html/index.html (1 line):
	- line 430: // display the images and the current predictions; TODO: should not color based on detection type 0 (ActiveDetection)--these are actually images flagged for labeling but may also have a machine predicted label


api/batch_processing/postprocessing/repeat_detection_elimination/repeat_detections_core.py (1 line):
	- line 725: # TODO: in the case where we're loading an existing set of FPs after manual filtering,


benchmark/model_eval_utils.py (1 line):
	- line 129: # TODO move detector_output_path specific code out so that this function evaluates only on classification results (confidences)


api/batch_processing/integration/eMammal/WPF-integration-app/eMammalMySQLOps.cs (1 line):
	- line 84: // TODO: add error checking return null on error


archive/detection/eval/analyze_sequence_detection_one_guess_per_sequence.py (1 line):
	- line 163: temp_labels[j] = True #TODO: this currently only works for oneclass?


visualization/visualize_db.py (1 line):
	- line 274: # TODO: optionally write html only for images where rendering succeeded


api/synchronous/api_core/animal_detection_classification_api/tf_classifer.py (1 line):
	- line 2: TODO - Copy this module from the classification folder once it has been refactored so that we do not keep


archive/detection/eval/analyze_sequence_detection.py (1 line):
	- line 163: temp_labels[j] = True #TODO: this currently only works for oneclass?


archive/index.html (1 line):
	- line 378: // display the images and the current predictions; TODO: should not color based on detection type 0 (ActiveDetection)--these are actually images flagged for labeling but may also have a machine predicted label


research/active_learning/DL/sqlite_data_loader.py (1 line):
	- line 94: # TODO: should this also change self.kind?


research/active_learning/Database/initialize_pretrain_db.py (1 line):
	- line 11: #  TODO update: Assumes that crops have already


archive/runapp.py (1 line):
	- line 73: args.strategy = 'confidence' ## TODO: hard coding this for now


archive/detection/eval/analyze_image_detection.py (1 line):
	- line 140: temp_labels[j] = True #TODO: this currently only works for oneclass?


api/batch_processing/postprocessing/load_api_results.py (1 line):
	- line 178: # TODO: hit some silly issues with vectorized str() and escaped characters, vectorize


research/active_learning/Database/initialize_target_db_from_classification_dataset.py (1 line):
	- line 72: ## TODO: allow user to update the class list through the labeling tool UI as they see different species


data_management/annotations/add_bounding_boxes_to_json.py (1 line):
	- line 150: 'location': location  # TODO bug - location should be region_name + camera_location


api/batch_processing/api_core/server_api_config.py (1 line):
	- line 50: # TODO add MD versions info to AppConfig


api/batch_processing/api_core/server_job_status_table.py (1 line):
	- line 70: # TODO do not read the entry first to get the call_params when the Cosmos SDK add a


api/batch_processing/integration/eMammal/WPF-integration-app/eMammalIntegrationWindow.xaml.cs (1 line):
	- line 275: // TODO: change this code copied from web


research/active_learning/run.py (1 line):
	- line 150: dataset_query = Detection.select(Detection.image_id, Oracle.label, Detection.kind).join(Oracle).order_by(fn.random()).limit(args.db_query_limit) ## TODO: should this really be order_by random?


data_management/importers/snapshot_safari_importer.py (1 line):
	- line 164: # TODO: iterrows() is a terrible way to do this, but this is one of those days


archive/detection/eval/analyze_image_detection_one_guess_per_image.py (1 line):
	- line 144: #    temp_labels[j] = True #TODO: this currently only works for oneclass?