megadetector 5.0.25__py3-none-any.whl → 5.0.27__py3-none-any.whl

megadetector/postprocessing/create_crop_folder.py

@@ -18,6 +18,7 @@ from collections import defaultdict
 from functools import partial
 
 from megadetector.utils.path_utils import insert_before_extension
+from megadetector.utils.ct_utils import invert_dictionary
 from megadetector.visualization.visualization_utils import crop_image
 from megadetector.visualization.visualization_utils import exif_preserving_save
 
@@ -48,6 +49,11 @@ class CreateCropFolderOptions:
         
         #: Whether to use processes ('process') or threads ('thread') for parallelization
         self.pool_type = 'thread'
+
+        #: Include only these categories, or None to include all
+        #:
+        #: options.category_names_to_include = ['animal']
+        self.category_names_to_include = None
                 
           
 #%% Support functions
@@ -106,7 +112,8 @@ def _generate_crops_for_single_image(crops_this_image,
 
 def crop_results_to_image_results(image_results_file_with_crop_ids,
                                   crop_results_file,
-                                  output_file):
+                                  output_file,
+                                  delete_crop_information=True):
     """
     This function is intended to be run after you have:
         
@@ -115,14 +122,18 @@ def crop_results_to_image_results(image_results_file_with_crop_ids,
         3. Run a species classifier on those crops
     
     This function will take the crop-level results and transform them back
-    to the original images.
+    to the original images.  Classification categories, if available, are taken 
+    from [crop_results_file].
         
     Args:
         image_results_file_with_crop_ids (str): results file for the original images,
-            containing crop IDs, likely generated via create_crop_folder.
+            containing crop IDs, likely generated via create_crop_folder.  All 
+            non-standard fields in this file will be passed along to [output_file].
         crop_results_file (str): results file for the crop folder
         output_file (str): ouptut .json file, containing crop-level classifications
-            mapped back to the image level
+            mapped back to the image level.
+        delete_crop_information (bool, optional): whether to delete the "crop_id" and
+            "crop_filename_relative" fields from each detection, if present.
     """
     
     ##%% Validate inputs
@@ -136,26 +147,45 @@ def crop_results_to_image_results(image_results_file_with_crop_ids,
     
     ##%% Read input files
     
+    print('Reading input...')
+    
     with open(image_results_file_with_crop_ids,'r') as f:
         image_results_with_crop_ids = json.load(f)
     with open(crop_results_file,'r') as f:
         crop_results = json.load(f)
-    
-    assert crop_results['detection_categories'] == \
-        image_results_with_crop_ids['detection_categories'], \
-            'Crop results and image-level results use different detection categories'
-        
+
+    # Find all the detection categories that need to be consistent
+    used_category_ids = set()
+    for im in tqdm(image_results_with_crop_ids['images']):
+        if 'detections' not in im or im['detections'] is None:
+            continue        
+        for det in im['detections']:
+            if 'crop_id' in det:
+                used_category_ids.add(det['category'])
+    
+    # Make sure the categories that matter are consistent across the two files
+    for category_id in used_category_ids:
+        category_name = image_results_with_crop_ids['detection_categories'][category_id]
+        assert category_id in crop_results['detection_categories'] and \
+            category_name == crop_results['detection_categories'][category_id], \
+                'Crop results and detection results use incompatible categories'
+                
     crop_filename_to_results = {}
     
     # im = crop_results['images'][0]
     for im in crop_results['images']:
         crop_filename_to_results[im['file']] = im
     
-    image_results_with_crop_ids['classification_categories'] = \
-        crop_results['classification_categories']
+    if 'classification_categories' in crop_results:
+        image_results_with_crop_ids['classification_categories'] = \
+            crop_results['classification_categories']
+            
+    if 'classification_category_descriptions' in crop_results:
+        image_results_with_crop_ids['classification_category_descriptions'] = \
+            crop_results['classification_category_descriptions']
     
         
-    ##%% Read classifications from crop results
+    ##%% Read classifications from crop results, merge into image-level results
     
     # im = image_results_with_crop_ids['images'][0]
     for im in tqdm(image_results_with_crop_ids['images']):
@@ -175,11 +205,18 @@ def crop_results_to_image_results(image_results_file_with_crop_ids,
                 crop_results_this_detection = crop_filename_to_results[crop_filename_relative]
                 assert crop_results_this_detection['file'] == crop_filename_relative
                 assert len(crop_results_this_detection['detections']) == 1
-                assert crop_results_this_detection['detections'][0]['conf'] == det['conf']
+                # Allow a slight confidence difference for the case where output precision was truncated
+                assert abs(crop_results_this_detection['detections'][0]['conf'] - det['conf']) < 0.01
                 assert crop_results_this_detection['detections'][0]['category'] == det['category']
                 assert crop_results_this_detection['detections'][0]['bbox'] == [0,0,1,1]
                 det['classifications'] = crop_results_this_detection['detections'][0]['classifications']
-                
+            
+            if delete_crop_information:
+                if 'crop_id' in det:
+                    del det['crop_id']
+                if 'crop_filename_relative' in det:
+                    del det['crop_filename_relative']
+                    
         # ...for each detection
     
     # ...for each image        
@@ -187,6 +224,8 @@ def crop_results_to_image_results(image_results_file_with_crop_ids,
     
     ##%% Write output file
     
+    print('Writing output file...')
+    
     with open(output_file,'w') as f:
         json.dump(image_results_with_crop_ids,f,indent=1)
 
@@ -223,21 +262,35 @@ def create_crop_folder(input_file,
     assert os.path.isfile(input_file), 'Input file {} not found'.format(input_file)
     assert os.path.isdir(input_folder), 'Input folder {} not found'.format(input_folder)
     os.makedirs(output_folder,exist_ok=True)
-    os.makedirs(os.path.dirname(output_file),exist_ok=True)
+    if output_file is not None:
+        os.makedirs(os.path.dirname(output_file),exist_ok=True)
     
     
     ##%% Read input
     
+    print('Reading MD results file...')    
     with open(input_file,'r') as f:
         detection_results = json.load(f)
-       
-        
-    ##%% Make a list crops that we need to create
+    
+    category_ids_to_include = None
+                
+    if options.category_names_to_include is not None:        
+        category_id_to_name = detection_results['detection_categories']
+        category_name_to_id = invert_dictionary(category_id_to_name)    
+        category_ids_to_include = set()
+        for category_name in options.category_names_to_include:
+            assert category_name in category_name_to_id, \
+                'Unrecognized category name {}'.format(category_name)
+            category_ids_to_include.add(category_name_to_id[category_name])        
+
+    ##%% Make a list of crops that we need to create
     
     # Maps input images to list of dicts, with keys 'crop_id','detection'
     image_fn_relative_to_crops = defaultdict(list)
     n_crops = 0
     
+    n_detections_excluded_by_category = 0
+
     # im = detection_results['images'][0]
     for i_image,im in enumerate(detection_results['images']):
         
@@ -249,27 +302,35 @@ def create_crop_folder(input_file,
         image_fn_relative = im['file']
         
         for i_detection,det in enumerate(detections_this_image):
+                        
+            if det['conf'] < options.confidence_threshold:
+                continue
+
+            if (category_ids_to_include is not None) and \
+                (det['category'] not in category_ids_to_include):
+                n_detections_excluded_by_category += 1
+                continue
+
+            det['crop_id'] = i_detection
             
-            if det['conf'] > options.confidence_threshold:
-                            
-                det['crop_id'] = i_detection
-                
-                crop_info = {'image_fn_relative':image_fn_relative,
-                             'crop_id':i_detection,
-                             'detection':det}
-                
-                crop_filename_relative = _get_crop_filename(image_fn_relative, 
-                                                            crop_info['crop_id'])
-                det['crop_filename_relative'] = crop_filename_relative
+            crop_info = {'image_fn_relative':image_fn_relative,
+                            'crop_id':i_detection,
+                            'detection':det}
+            
+            crop_filename_relative = _get_crop_filename(image_fn_relative, 
+                                                        crop_info['crop_id'])
+            det['crop_filename_relative'] = crop_filename_relative
 
-                image_fn_relative_to_crops[image_fn_relative].append(crop_info)
-                n_crops += 1
-                
+            image_fn_relative_to_crops[image_fn_relative].append(crop_info)
+            n_crops += 1
+            
     # ...for each input image   
 
     print('Prepared a list of {} crops from {} of {} input images'.format(
         n_crops,len(image_fn_relative_to_crops),len(detection_results['images'])))
-        
+    
+    if n_detections_excluded_by_category > 0:
+        print('Excluded {} detections by category'.format(n_detections_excluded_by_category))
     
     ##%% Generate crops
         
@@ -307,6 +368,7 @@ def create_crop_folder(input_file,
 
     # ...if we're using parallel processing    
     
+    
     ##%% Write output file
     
     if output_file is not None:

megadetector/postprocessing/load_api_results.py

@@ -107,6 +107,9 @@ def write_api_results(detection_results_table, other_fields, out_path):
     images = detection_results_table.to_json(orient='records',
                                              double_precision=3)
     images = json.loads(images)
+    for im in images:
+        if 'failure' in im and im['failure'] is None:
+            del im['failure']
     fields['images'] = images
     
     # Convert the 'version' field back to a string as per format convention
@@ -129,7 +132,7 @@ def write_api_results(detection_results_table, other_fields, out_path):
     except Exception:
         print('Warning: error removing max_detection_conf from output')
         pass
-
+    
     with open(out_path, 'w') as f:
         json.dump(fields, f, indent=1)
 

megadetector/postprocessing/md_to_coco.py

@@ -181,7 +181,7 @@ def md_to_coco(md_results_file,
             
             w = im['width']
             h = im['height']
-            
+        
         coco_im['width'] = w
         coco_im['height'] = h
         

megadetector/postprocessing/postprocess_batch_results.py

@@ -30,6 +30,7 @@ import time
 import uuid
 import warnings
 import random
+import json
 
 from enum import IntEnum
 from multiprocessing.pool import ThreadPool
@@ -48,8 +49,11 @@ from megadetector.visualization import visualization_utils as vis_utils
 from megadetector.visualization import plot_utils
 from megadetector.utils.write_html_image_list import write_html_image_list
 from megadetector.utils import path_utils
-from megadetector.utils.ct_utils import args_to_object, sets_overlap
-from megadetector.data_management.cct_json_utils import (CameraTrapJsonUtils, IndexedJsonDb)
+from megadetector.utils.ct_utils import args_to_object
+from megadetector.utils.ct_utils import sets_overlap
+from megadetector.utils.ct_utils import sort_dictionary_by_value
+from megadetector.data_management.cct_json_utils import CameraTrapJsonUtils
+from megadetector.data_management.cct_json_utils import IndexedJsonDb
 from megadetector.postprocessing.load_api_results import load_api_results
 from megadetector.detection.run_detector import get_typical_confidence_threshold_from_results
 
@@ -214,6 +218,15 @@ class PostProcessingOptions:
         #: Character encoding to use when writing the index HTML html
         self.output_html_encoding = None
         
+        #: Additional image fields to display in image headers.  If this is a list,
+        #: we'll include those fields; if this is a dict, we'll use that dict to choose
+        #: alternative display names for each field.
+        self.additional_image_fields_to_display = None
+        
+        #: If classification results are present, should we include a summary of 
+        #: classification categories?
+        self.include_classification_category_report = True
+        
     # ...__init__()
     
 # ...PostProcessingOptions
@@ -434,15 +447,6 @@ def _render_bounding_boxes(
     if options is None:
         options = PostProcessingOptions()
 
-    # Leaving code in place for reading from blob storage, may support this
-    # in the future.
-    """
-    stream = io.BytesIO()
-    _ = blob_service.get_blob_to_stream(container_name, image_id, stream)
-    # resize is to display them in this notebook or in the HTML more quickly
-    image = Image.open(stream).resize(viz_size)
-    """
-
     image_full_path = None
     
     if res in options.rendering_bypass_sets:
@@ -472,10 +476,12 @@ def _render_bounding_boxes(
         if image is not None:
             
             original_size = image.size
-            
+    
+            # Resize the image if necessary
             if options.viz_target_width is not None:
                 image = vis_utils.resize_image(image, options.viz_target_width)
     
+            # Render ground truth boxes if necessary
             if ground_truth_boxes is not None and len(ground_truth_boxes) > 0:
                 
                 # Create class labels like "gt_1" or "gt_27"
@@ -487,8 +493,7 @@ def _render_bounding_boxes(
                                                    original_size=original_size,label_map=label_map,
                                                    thickness=4,expansion=4)
         
-            # render_detection_bounding_boxes expects either a float or a dict mapping
-            # category IDs to names.
+            # Preprare per-category confidence thresholds
             if isinstance(options.confidence_threshold,float):
                 rendering_confidence_threshold = options.confidence_threshold
             else:
@@ -499,12 +504,14 @@ def _render_bounding_boxes(
                 for category_id in category_ids:
                     rendering_confidence_threshold[category_id] = \
                         _get_threshold_for_category_id(category_id, options, detection_categories)
-                
+            
+            # Render detection boxes
             vis_utils.render_detection_bounding_boxes(
                 detections, image,
                 label_map=detection_categories,
                 classification_label_map=classification_categories,
                 confidence_threshold=rendering_confidence_threshold,
+                classification_confidence_threshold=options.classification_confidence_threshold,
                 thickness=options.line_thickness,
                 expansion=options.box_expansion)
     
@@ -686,9 +693,11 @@ def _has_positive_detection(detections,options,detection_categories):
     return found_positive_detection
         
 
-def _render_image_no_gt(file_info,detection_categories_to_results_name,
-                       detection_categories,classification_categories,
-                       options):
+def _render_image_no_gt(file_info,
+                        detection_categories_to_results_name,
+                        detection_categories,
+                        classification_categories,
+                        options):
     """
     Renders an image (with no ground truth information)
     
@@ -713,9 +722,15 @@ def _render_image_no_gt(file_info,detection_categories_to_results_name,
     Returns None if there are any errors.
     """
     
-    image_relative_path = file_info[0]
-    max_conf = file_info[1]
-    detections = file_info[2]
+    image_relative_path = file_info['file']
+    
+    # Useful debug snippet
+    #
+    # if 'filename' in image_relative_path:
+    #    import pdb; pdb.set_trace()
+    
+    max_conf = file_info['max_detection_conf']
+    detections = file_info['detections']
 
     # Determine whether any positive detections are present (using a threshold that
     # may vary by category)
@@ -749,9 +764,31 @@ def _render_image_no_gt(file_info,detection_categories_to_results_name,
         assert detection_status == DetectionStatus.DS_ALMOST
         res = 'almost_detections'
 
-    display_name = '<b>Result type</b>: {}, <b>Image</b>: {}, <b>Max conf</b>: {:0.3f}'.format(
+    display_name = '<b>Result type</b>: {}, <b>image</b>: {}, <b>max conf</b>: {:0.3f}'.format(
         res, image_relative_path, max_conf)
 
+    # Are there any bonus fields we need to include in each image header?
+    if options.additional_image_fields_to_display is not None:
+        
+        for field_name in options.additional_image_fields_to_display:
+            
+            if field_name in file_info:
+                
+                field_value = file_info[field_name]
+                
+                if (field_value is None) or \
+                    (isinstance(field_value,float) and np.isnan(field_value)):
+                        continue
+                
+                # Optionally use a display name that's different from the field name
+                if isinstance(options.additional_image_fields_to_display,dict):
+                    field_display_name = \
+                        options.additional_image_fields_to_display[field_name]
+                else:
+                    field_display_name = field_name
+                field_string = '<b>{}</b>: {}'.format(field_display_name,field_value)
+                display_name += ', {}'.format(field_string)
+                
     rendering_options = copy.copy(options)
     if detection_status == DetectionStatus.DS_ALMOST:
         rendering_options.confidence_threshold = \
@@ -781,17 +818,24 @@ def _render_image_no_gt(file_info,detection_categories_to_results_name,
             if det['conf'] > max_conf:
                 max_conf = det['conf']
 
+            # We make the decision here that only "detections" (not "almost-detections")
+            # will appear on the classification category pages
+            detection_threshold = \
+                _get_threshold_for_category_id(det['category'], options, detection_categories)
+            if det['conf'] < detection_threshold:
+                continue
+                
             if ('classifications' in det) and (len(det['classifications']) > 0) and \
                 (res != 'non_detections'):
 
-                # This is a list of [class,confidence] pairs, sorted by confidence
+                # This is a list of [class,confidence] pairs, sorted by classification confidence
                 classifications = det['classifications']
                 top1_class_id = classifications[0][0]
                 top1_class_name = classification_categories[top1_class_id]
                 top1_class_score = classifications[0][1]
 
-                # If we either don't have a confidence threshold, or we've met our
-                # confidence threshold
+                # If we either don't have a classification confidence threshold, or 
+                # we've met our classification confidence threshold
                 if (options.classification_confidence_threshold < 0) or \
                     (top1_class_score >= options.classification_confidence_threshold):
                     class_string = 'class_{}'.format(top1_class_name)                    
@@ -823,9 +867,9 @@ def _render_image_with_gt(file_info,ground_truth_indexed_db,
     data format.
     """
     
-    image_relative_path = file_info[0]
-    max_conf = file_info[1]
-    detections = file_info[2]
+    image_relative_path = file_info['file']
+    max_conf = file_info['max_detection_conf']
+    detections = file_info['detections']
 
     # This should already have been normalized to either '/' or '\'
 
@@ -971,6 +1015,7 @@ def process_batch_results(options):
             print('\n*** Warning: {} images with ambiguous positive/negative status found in ground truth ***\n'.format(
                 n_ambiguous))
 
+
     ##%% Load detection (and possibly classification) results
 
     # If the caller hasn't supplied results, load them
@@ -1028,6 +1073,8 @@ def process_batch_results(options):
     n_positives = 0
     n_almosts = 0
     
+    print('Assigning images to rendering categories')
+    
     for i_row,row in tqdm(detections_df.iterrows(),total=len(detections_df)):
         
         detections = row['detections']
@@ -1372,7 +1419,7 @@ def process_batch_results(options):
         for _, row in images_to_visualize.iterrows():
 
             # Filenames should already have been normalized to either '/' or '\'
-            files_to_render.append([row['file'], row['max_detection_conf'], row['detections']])
+            files_to_render.append(row.to_dict())
 
         start_time = time.time()
         if options.parallelize_rendering:
@@ -1523,8 +1570,13 @@ def process_batch_results(options):
                     len(images_html['class_{}'.format(cname)]))
             index_page += '</div>'
 
-        # Close body and html tags
-        index_page += '{}</body></html>'.format(options.footer_text)
+        # Write custom footer if it was provided
+        if (options.footer_text is not None) and (len(options.footer_text) > 0):
+            index_page += '{}\n'.format(options.footer_text)
+        
+        # Close open html tags
+        index_page += '\n</body></html>\n'
+        
         output_html_file = os.path.join(output_dir, 'index.html')
         with open(output_html_file, 'w', 
                   encoding=options.output_html_encoding) as f:
@@ -1532,7 +1584,7 @@ def process_batch_results(options):
 
         print('Finished writing html to {}'.format(output_html_file))
 
-    # ...for each image
+    # ...if we have ground truth
 
 
     ##%% Otherwise, if we don't have ground truth...
@@ -1618,9 +1670,7 @@ def process_batch_results(options):
             assert isinstance(row['detections'],list)
             
             # Filenames should already have been normalized to either '/' or '\'
-            files_to_render.append([row['file'],
-                                    row['max_detection_conf'],
-                                    row['detections']])
+            files_to_render.append(row.to_dict())
 
         start_time = time.time()
         if options.parallelize_rendering:
@@ -1691,8 +1741,7 @@ def process_batch_results(options):
         # Write index.html
 
         # We can't just sum these, because image_counts includes images in both their
-        # detection and classification classes
-        # total_images = sum(image_counts.values())
+        # detection and classification classes        
         total_images = 0
         for k in image_counts.keys():
             v = image_counts[k]
@@ -1779,9 +1828,15 @@ def process_batch_results(options):
             else:
                 index_page += '<a href="{}">{}</a> ({}, {:.1%})<br/>\n'.format(
                     filename,label,image_count,image_fraction)
-
+        
+        # ...for each result set
+        
         index_page += '</div>\n'
 
+        # If classification information is present and we're supposed to create
+        # a summary of classifications, we'll put it here
+        category_count_footer = None
+        
         if has_classification_info:
             
             index_page += '<h3>Species classification results</h3>'
@@ -1810,15 +1865,74 @@ def process_batch_results(options):
                         cname, cname.lower(), ccount)
             index_page += '</div>\n'
 
-        index_page += '{}</body></html>'.format(options.footer_text)
+            if options.include_classification_category_report:
+
+                # TODO: it's only for silly historical reasons that we re-read
+                # the input file in this case; we're not currently carrying the json
+                # representation around, only the Pandas representation.
+                
+                print('Generating classification category report')
+                
+                with open(options.md_results_file,'r') as f:
+                    d = json.load(f)
+                
+                classification_category_to_count = {}
+
+                # im = d['images'][0]
+                for im in d['images']:
+                    if 'detections' in im and im['detections'] is not None:
+                        for det in im['detections']:
+                            if 'classifications' in det:
+                                class_id = det['classifications'][0][0]
+                                if class_id not in classification_category_to_count:
+                                    classification_category_to_count[class_id] = 0
+                                else:
+                                    classification_category_to_count[class_id] = \
+                                        classification_category_to_count[class_id] + 1
+
+                category_name_to_count = {}
+
+                for class_id in classification_category_to_count:
+                    category_name = d['classification_categories'][class_id]
+                    category_name_to_count[category_name] = \
+                        classification_category_to_count[class_id]
+
+                category_name_to_count = sort_dictionary_by_value(
+                    category_name_to_count,reverse=True)
+
+                category_count_footer = ''
+                category_count_footer += '<br/>\n'
+                category_count_footer += \
+                    '<h3>Category counts (for the whole dataset, not just the sample used for this page)</h3>\n'
+                category_count_footer += '<div class="contentdiv">\n'
+
+                for category_name in category_name_to_count.keys():
+                    count = category_name_to_count[category_name]
+                    category_count_html = '{}: {}<br>\n'.format(category_name,count)    
+                    category_count_footer += category_count_html
+
+                category_count_footer += '</div>\n'
+                
+            # ...if we're generating a classification category report
+            
+        # ...if classification info is present
+        
+        if category_count_footer is not None:
+            index_page += category_count_footer + '\n'
+            
+        # Write custom footer if it was provided
+        if (options.footer_text is not None) and (len(options.footer_text) > 0):
+            index_page += options.footer_text + '\n'
+        
+        # Close open html tags
+        index_page += '\n</body></html>\n'
+        
         output_html_file = os.path.join(output_dir, 'index.html')
         with open(output_html_file, 'w', 
                   encoding=options.output_html_encoding) as f:
             f.write(index_page)
 
-        print('Finished writing html to {}'.format(output_html_file))
-
-        # os.startfile(output_html_file)
+        print('Finished writing html to {}'.format(output_html_file))        
 
     # ...if we do/don't have ground truth
 

megadetector/postprocessing/repeat_detection_elimination/find_repeat_detections.py

@@ -41,7 +41,6 @@ if False:
     baseDir = ''
 
     options = repeat_detections_core.RepeatDetectionOptions()
-    options.bRenderHtml = True
     options.imageBase = baseDir
     options.outputBase = os.path.join(baseDir, 'repeat_detections')
     options.filenameReplacements = {}  # E.g., {'20190430cameratraps\\':''}
@@ -85,11 +84,10 @@ def main():
                              'do manual review of the repeat detection images (which you should)')
         
     parser.add_argument('--imageBase', action='store', type=str, default='',
-                        help='Image base dir, relevant if renderHtml is True or if ' + \
-                             '"omitFilteringFolder" is not set')
+                        help='Image base dir')
                             
     parser.add_argument('--outputBase', action='store', type=str, default='',
-                        help='HTML or filtering folder output dir')
+                        help='filtering folder output dir')
     
     parser.add_argument('--confidenceMin', action='store', type=float,
                         default=defaultOptions.confidenceMin,
@@ -146,7 +144,7 @@ def main():
 
     parser.add_argument('--omitFilteringFolder', action='store_false',
                         dest='bWriteFilteringFolder',
-                        help='Should we create a folder of rendered detections for post-filtering?')
+                        help='Should we skip creating the folder of rendered detections filtering?')
     
     parser.add_argument('--debugMaxDir', action='store', type=int, default=-1, 
                         help='For debugging only, limit the number of directories we process')
@@ -191,9 +189,6 @@ def main():
                         default=defaultOptions.detectionTilesPrimaryImageWidth,
                         help='The width of the main image when rendering images with detection tiles')
 
-    parser.add_argument('--renderHtml', action='store_true',
-                        dest='bRenderHtml', help='Should we render HTML output?')
-    
     if len(sys.argv[1:]) == 0:
         parser.print_help()
         parser.exit()