megadetector 5.0.15__py3-none-any.whl → 5.0.17__py3-none-any.whl

megadetector/postprocessing/md_to_coco.py

@@ -31,11 +31,18 @@ def md_to_coco(md_results_file,
                validate_image_sizes=False,
                info=None,
                preserve_nonstandard_metadata=True,
-               include_failed_images=True):
+               include_failed_images=True,
+               include_annotations_without_bounding_boxes=True,
+               empty_category_id='0'):
     """
     "Converts" MegaDetector output files to COCO format.  "Converts" is in quotes because
     this is an opinionated transformation that requires a confidence threshold.
     
+    The default confidence threshold is not 0; the assumption is that by default, you are 
+    going to treat the resulting COCO file as a set of labels.  If you are using the resulting COCO
+    file to evaluate a detector, you likely want a default confidence threshold of 0.  Confidence
+    values will be written to the semi-standard "score" field for each image
+    
     A folder of images is required if width and height information are not available 
     in the MD results file.
 
@@ -54,8 +61,13 @@ def md_to_coco(md_results_file,
         preserve_nonstandard_metadata (bool, optional): if this is True, confidence will be preserved in a 
             non-standard "conf" field in each annotation, and any random fields present in each image's data
             (e.g. EXIF metadata) will be propagated to COCO output    
-        include_failed_images (boo, optional): if this is True, failed images will be propagated to COCO output 
+        include_failed_images (bool, optional): if this is True, failed images will be propagated to COCO output 
             with a non-empty "failure" field and no other fields, otherwise failed images will be skipped.
+        include_annotations_without_bounding_boxes (bool, optional): if this is True, annotations with
+            only class labels (no bounding boxes) will be included in the output.  If this is False, empty
+            images will be represented with no annotations.
+        empty_category_id (str, optional): category ID reserved for the 'empty' class, should not be
+            attached to any bounding boxes
     
     Returns:
         dict: the COCO data dict, identical to what's written to [coco_output_file] if [coco_output_file]
@@ -67,6 +79,8 @@ def md_to_coco(md_results_file,
         
     coco_images = []
     coco_annotations = []
+        
+    print('Converting MD results to COCO...')
     
     # im = md_results['images'][0]
     for im in tqdm(md_results['images']):
@@ -129,13 +143,13 @@ def md_to_coco(md_results_file,
             coco_category_id = int(md_category_id)
             ann['category_id'] = coco_category_id
             
-            # In very esoteric cases, we use the empty category (0) in MD-formatted output files
-            if md_category_id != '0':
+            if md_category_id != empty_category_id:
                 
                 assert 'bbox' in detection,\
                     'Oops: non-empty category with no bbox in {}'.format(im['file'])
             
                 ann['bbox'] = detection['bbox']
+                
                 # MegaDetector: [x,y,width,height] (normalized, origin upper-left)
                 # COCO: [x,y,width,height] (absolute, origin upper-left)
                 ann['bbox'][0] = ann['bbox'][0] * coco_im['width']
@@ -144,13 +158,19 @@ def md_to_coco(md_results_file,
                 ann['bbox'][3] = ann['bbox'][3] * coco_im['height']                
             
             else:
-                
-                print('Warning: empty category annotation in file {}'.format(im['file']))
+            
+                # In very esoteric cases, we use the empty category (0) in MD-formatted output files
+                print('Warning: empty category ({}) used for annotation in file {}'.format(
+                    empty_category_id,im['file']))
+                pass
                       
             if preserve_nonstandard_metadata:
-                ann['conf'] = detection['conf']
-                
-            coco_annotations.append(ann)            
+                # "Score" is a semi-standard string here, recognized by at least pycocotools
+                # ann['conf'] = detection['conf']
+                ann['score'] = detection['conf']
+        
+            if 'bbox' in ann or include_annotations_without_bounding_boxes:
+                coco_annotations.append(ann)            
         
         # ...for each detection
     
@@ -176,6 +196,8 @@ def md_to_coco(md_results_file,
                          'name':md_results['detection_categories'][md_category_id]}
         output_dict['categories'].append(coco_category)
         
+    print('Writing COCO output file...')
+    
     if coco_output_file is not None:
         with open(coco_output_file,'w') as f:
             json.dump(output_dict,f,indent=1)

megadetector/postprocessing/postprocess_batch_results.py

@@ -770,7 +770,7 @@ def _render_image_no_gt(file_info,detection_categories_to_results_name,
             if det['conf'] > max_conf:
                 max_conf = det['conf']
 
-            if ('classifications' in det):
+            if ('classifications' in det) and (len(det['classifications']) > 0):
 
                 # This is a list of [class,confidence] pairs, sorted by confidence
                 classifications = det['classifications']
@@ -949,6 +949,13 @@ def process_batch_results(options):
               f'negative, {n_positive} positive, {n_unknown} unknown, '
               f'{n_ambiguous} ambiguous')
 
+        if n_positive == 0:
+            print('\n*** Warning: no positives found in ground truth, analysis won\'t be very meaningful ***\n')
+        if n_negative == 0:
+            print('\n*** Warning: no negatives found in ground truth, analysis won\'t be very meaningful ***\n')
+        if n_ambiguous > 0:
+            print('\n*** Warning: {} images with ambiguous positive/negative status found in ground truth ***\n'.format(
+                n_ambiguous))
 
     ##%% Load detection (and possibly classification) results
 
@@ -1095,25 +1102,34 @@ def process_batch_results(options):
 
         ##%% Detection evaluation: compute precision/recall
 
-        # numpy array of detection probabilities
+        # numpy array of maximum confidence values
         p_detection = detections_df['max_detection_conf'].values
-        n_detections = len(p_detection)
+        n_detection_values = len(p_detection)
 
         # numpy array of bools (0.0/1.0), and -1 as null value
-        gt_detections = np.zeros(n_detections, dtype=float)
+        gt_detections = np.zeros(n_detection_values, dtype=float)
 
+        n_positive = 0
+        n_negative = 0
+        
         for i_detection, fn in enumerate(detector_files):
+            
             image_id = ground_truth_indexed_db.filename_to_id[fn]
             image = ground_truth_indexed_db.image_id_to_image[image_id]
             detection_status = image['_detection_status']
 
             if detection_status == DetectionStatus.DS_NEGATIVE:
                 gt_detections[i_detection] = 0.0
+                n_negative += 1
             elif detection_status == DetectionStatus.DS_POSITIVE:
                 gt_detections[i_detection] = 1.0
+                n_positive += 1
             else:
                 gt_detections[i_detection] = -1.0
 
+        print('Of {} ground truth values, found {} positives and {} negatives'.format(
+            len(detections_df),n_positive,n_negative))
+        
         # Don't include ambiguous/unknown ground truth in precision/recall analysis
         b_valid_ground_truth = gt_detections >= 0.0
 
@@ -1187,13 +1203,13 @@ def process_batch_results(options):
         # Rows / first index is ground truth, columns / second index is predicted category
         classifier_cm = collections.defaultdict(lambda: collections.defaultdict(lambda: 0))
 
-        # iDetection = 0; fn = detector_files[iDetection]; print(fn)
+        # i_detection = 0; fn = detector_files[i_detection]; print(fn)
         assert len(detector_files) == len(detections_df)
-        for iDetection, fn in enumerate(detector_files):
+        for i_detection, fn in enumerate(detector_files):
 
             image_id = ground_truth_indexed_db.filename_to_id[fn]
             image = ground_truth_indexed_db.image_id_to_image[image_id]
-            detections = detections_df['detections'].iloc[iDetection]
+            detections = detections_df['detections'].iloc[i_detection]
             pred_class_ids = [det['classifications'][0][0] \
                 for det in detections if 'classifications' in det.keys()]
             pred_classnames = [classification_categories[pd] for pd in pred_class_ids]

megadetector/postprocessing/repeat_detection_elimination/repeat_detections_core.py

@@ -210,9 +210,12 @@ class RepeatDetectionOptions:
         #: a/b/c/RECONYX100 and a/b/c/RECONYX101 may really be the same camera).
         #:
         #: See ct_utils for a common replacement function that handles most common
-        #: manufacturer folder names.
+        #: manufacturer folder names:
+        #:
+        #: from megadetector.utils import ct_utils
+        #: self.customDirNameFunction = ct_utils.image_file_to_camera_folder
         self.customDirNameFunction = None
-        
+                
         #: Include only specific folders, mutually exclusive with [excludeFolders]
         self.includeFolders = None
         

megadetector/postprocessing/subset_json_detector_output.py

@@ -124,7 +124,7 @@ class SubsetJsonDetectorOutputOptions:
         self.remove_failed_images = False
         
         #: Either a list of category IDs (as string-ints) (not names), or a dictionary mapping category *IDs* 
-        #: (as string-ints) (not names) to thresholds.  Removes  non-matching detections, does not 
+        #: (as string-ints) (not names) to thresholds.  Removes non-matching detections, does not 
         #: remove images.  Not technically mutually exclusize with category_names_to_keep, but it's an esoteric 
         #: scenario indeed where you would want to specify both.
         self.categories_to_keep = None
@@ -517,7 +517,7 @@ def subset_json_detector_output(input_filename, output_filename, options, data=N
     else:
         
         # Map images to unique folders
-        print('Finding unique folders')    
+        print('Finding unique folders')
         
         folders_to_images = {}
         
@@ -670,16 +670,26 @@ def main():
     parser = argparse.ArgumentParser()
     parser.add_argument('input_file', type=str, help='Input .json filename')
     parser.add_argument('output_file', type=str, help='Output .json filename')
-    parser.add_argument('--query', type=str, default=None, help='Query string to search for (omitting this matches all)')
-    parser.add_argument('--replacement', type=str, default=None, help='Replace [query] with this')
-    parser.add_argument('--confidence_threshold', type=float, default=None, help='Remove detections below this confidence level')
-    parser.add_argument('--split_folders', action='store_true', help='Split .json files by leaf-node folder')
-    parser.add_argument('--split_folder_param', type=int, help='Directory level count for n_from_bottom and n_from_top splitting')
-    parser.add_argument('--split_folder_mode', type=str, help='Folder level to use for splitting ("top" or "bottom")')
-    parser.add_argument('--make_folder_relative', action='store_true', help='Make image paths relative to their containing folder (only meaningful with split_folders)')
-    parser.add_argument('--overwrite_json_files', action='store_true', help='Overwrite output files')
-    parser.add_argument('--copy_jsons_to_folders', action='store_true', help='When using split_folders and make_folder_relative, copy jsons to their corresponding folders (relative to output_file)')
-    parser.add_argument('--create_folders', action='store_true', help='When using copy_jsons_to_folders, create folders that don''t exist')    
+    parser.add_argument('--query', type=str, default=None, 
+                        help='Query string to search for (omitting this matches all)')
+    parser.add_argument('--replacement', type=str, default=None, 
+                        help='Replace [query] with this')
+    parser.add_argument('--confidence_threshold', type=float, default=None, 
+                        help='Remove detections below this confidence level')
+    parser.add_argument('--split_folders', action='store_true', 
+                        help='Split .json files by leaf-node folder')
+    parser.add_argument('--split_folder_param', type=int,
+                        help='Directory level count for n_from_bottom and n_from_top splitting')
+    parser.add_argument('--split_folder_mode', type=str,
+                        help='Folder level to use for splitting ("top" or "bottom")')
+    parser.add_argument('--make_folder_relative', action='store_true', 
+                        help='Make image paths relative to their containing folder (only meaningful with split_folders)')
+    parser.add_argument('--overwrite_json_files', action='store_true', 
+                        help='Overwrite output files')
+    parser.add_argument('--copy_jsons_to_folders', action='store_true', 
+                        help='When using split_folders and make_folder_relative, copy jsons to their corresponding folders (relative to output_file)')
+    parser.add_argument('--create_folders', action='store_true',
+                        help='When using copy_jsons_to_folders, create folders that don''t exist')    
     
     if len(sys.argv[1:]) == 0:
         parser.print_help()

megadetector/taxonomy_mapping/map_new_lila_datasets.py

@@ -15,10 +15,10 @@ import json
 # Created by get_lila_category_list.py
 input_lila_category_list_file = os.path.expanduser('~/lila/lila_categories_list/lila_dataset_to_categories.json')
 
-output_file = os.path.expanduser('~/lila/lila_additions_2023.12.29.csv')
+output_file = os.path.expanduser('~/lila/lila_additions_2024.07.16.csv')
 
 datasets_to_map = [
-    'Trail Camera Images of New Zealand Animals'
+    'Desert Lion Conservation Camera Traps'
     ]
 
 
@@ -133,7 +133,7 @@ if False:
     # q = 'white-throated monkey'
     # q = 'cingulata'
     # q = 'notamacropus'
-    q = 'porzana'
+    q = 'aves'
     taxonomy_preference = 'inat'
     m = get_preferred_taxonomic_match(q,taxonomy_preference)
     # print(m.scientific_name); import clipboard; clipboard.copy(m.scientific_name)

megadetector/taxonomy_mapping/prepare_lila_taxonomy_release.py

@@ -24,7 +24,7 @@ if False:
     release_taxonomy_file = os.path.expanduser('~/lila/lila-taxonomy-mapping_release.csv')
     # import clipboard; clipboard.copy(release_taxonomy_file)
 
-    # Created by get_lila_category_list.py... contains counts for each category
+    # Created by get_lila_annotation_counts.py... contains counts for each category
     lila_dataset_to_categories_file = os.path.expanduser('~/lila/lila_categories_list/lila_dataset_to_categories.json')
 
     assert os.path.isfile(lila_dataset_to_categories_file)
@@ -140,3 +140,4 @@ if False:
 
     print('Wrote final output to {}'.format(release_taxonomy_file))
 
+# ...if False

megadetector/taxonomy_mapping/preview_lila_taxonomy.py

@@ -16,7 +16,7 @@ import os
 import pandas as pd
 
 # lila_taxonomy_file = r"c:\git\agentmorrisprivate\lila-taxonomy\lila-taxonomy-mapping.csv"
-lila_taxonomy_file = os.path.expanduser('~/lila/lila_additions_2023.12.29.csv')
+lila_taxonomy_file = os.path.expanduser('~/lila/lila_additions_2024.07.16.csv')
 
 preview_base = os.path.expanduser('~/lila/lila_taxonomy_preview')
 os.makedirs(preview_base,exist_ok=True)

megadetector/taxonomy_mapping/simple_image_download.py

@@ -8,6 +8,11 @@ Slightly modified from:
 
 https://github.com/RiddlerQ/simple_image_download
 
+pip install python-magic
+
+# On Windows, also run:
+pip install python-magic-bin
+
 """
 
 #%% Imports

megadetector/taxonomy_mapping/species_lookup.py

@@ -208,7 +208,7 @@ def initialize_taxonomy_lookup(force_init=False) -> None:
     # Load GBIF taxonomy
     gbif_taxonomy_file = os.path.join(taxonomy_download_dir, 'GBIF', 'Taxon.tsv')
     print('Loading GBIF taxonomy from {}'.format(gbif_taxonomy_file))
-    gbif_taxonomy = pd.read_csv(gbif_taxonomy_file, sep='\t')
+    gbif_taxonomy = pd.read_csv(gbif_taxonomy_file, sep='\t', encoding='utf-8',on_bad_lines='warn')
     gbif_taxonomy['scientificName'] = gbif_taxonomy['scientificName'].fillna('').str.strip()
     gbif_taxonomy['canonicalName'] = gbif_taxonomy['canonicalName'].fillna('').str.strip()
 

megadetector/utils/ct_utils.py

@@ -16,6 +16,8 @@ import os
 import jsonpickle
 import numpy as np
 
+from operator import itemgetter
+
 # List of file extensions we'll consider images; comparisons will be case-insensitive
 # (i.e., no need to include both .jpg and .JPG on this list).
 image_extensions = ['.jpg', '.jpeg', '.gif', '.png']
@@ -294,6 +296,29 @@ def get_max_conf(im):
     return max_conf
 
 
+def sort_results_for_image(im):
+    """
+    Sort classification and detection results in descending order by confidence (in place).
+    
+    Args:
+        im (dict): image dictionary in the MD output format (with a 'detections' field)
+    """
+    if 'detections' not in im or im['detections'] is None:
+        return
+
+    # Sort detections in descending order by confidence
+    im['detections'] = sort_list_of_dicts_by_key(im['detections'],k='conf',reverse=True)
+    
+    for det in im['detections']:
+        
+        # Sort classifications (which are (class,conf) tuples) in descending order by confidence
+        if 'classifications' in det and \
+            (det['classifications'] is not None) and \
+            (len(det['classifications']) > 0):
+            L = det['classifications']
+            det['classifications'] = sorted(L,key=itemgetter(1),reverse=True)
+
+
 def point_dist(p1,p2):
     """
     Computes the distance between two points, represented as length-two tuples.
@@ -406,6 +431,21 @@ def split_list_into_n_chunks(L, n, chunk_strategy='greedy'):
         raise ValueError('Invalid chunk strategy: {}'.format(chunk_strategy))
 
 
+def sort_list_of_dicts_by_key(L,k,reverse=False):
+    """
+    Sorts the list of dictionaries [L] by the key [k].
+    
+    Args:
+        L (list): list of dictionaries to sort
+        k (object, typically str): the sort key
+        reverse (bool, optional): whether to sort in reverse (descending) order
+        
+    Returns:
+        dict: sorted copy of [d]
+    """
+    return sorted(L, key=lambda d: d[k], reverse=reverse)
+    
+    
 def sort_dictionary_by_key(d,reverse=False):
     """
     Sorts the dictionary [d] by key.
@@ -611,3 +651,11 @@ if False:
     r1 = [0.4,0.8,10,22]; r2 = [100, 101, 200, 210.4]; assert abs(rect_distance(r1,r2)-119.753) < 0.001
     r1 = [0.4,0.8,10,22]; r2 = [101, 101, 200, 210.4]; assert abs(rect_distance(r1,r2)-120.507) < 0.001    
     r1 = [0.4,0.8,10,22]; r2 = [120, 120, 200, 210.4]; assert abs(rect_distance(r1,r2)-147.323) < 0.001
+
+    
+    #%% Test dictionary sorting
+    
+    L = [{'a':5},{'a':0},{'a':10}]
+    k = 'a'
+    sort_list_of_dicts_by_key(L, k, reverse=True)
+