megadetector 5.0.6__py3-none-any.whl → 5.0.8__py3-none-any.whl

data_management/yolo_to_coco.py

@@ -2,10 +2,7 @@
 #
 # yolo_to_coco.py
 #
-# Converts a YOLO-formatted dataset to a COCO-formatted dataset. 
-#
-# Currently supports only a single folder (i.e., no recursion).  Treats images without
-# corresponding .txt files as empty.
+# Converts a folder of YOLO-formatted annotation files to a COCO-formatted dataset. 
 #
 ########
 
@@ -14,57 +11,241 @@
 import json
 import os
 
-from PIL import Image
+from multiprocessing.pool import ThreadPool
+from multiprocessing.pool import Pool
+from functools import partial
+
 from tqdm import tqdm
 
 from md_utils.path_utils import find_images
+from md_utils.ct_utils import invert_dictionary
+from md_visualization.visualization_utils import open_image
+from data_management.yolo_output_to_md_output import read_classes_from_yolo_dataset_file
+
+
+#%% Support functions
+
+def filename_to_image_id(fn):
+    return fn.replace(' ','_')
+
+def _process_image(fn_abs,input_folder,category_id_to_name):
+    """
+    Internal support function for processing one image's labels.
+    """
+    
+    # Create the image object for this image
+    fn_relative = os.path.relpath(fn_abs,input_folder)
+    image_id = filename_to_image_id(fn_relative)
+    
+    # This is done in a separate loop now
+    #
+    # assert image_id not in image_ids, \
+    #    'Oops, you have hit a very esoteric case where you have the same filename ' + \
+    #    'with both spaces and underscores, this is not currently handled.'
+    # image_ids.add(image_id)
+    
+    im = {}
+    im['file_name'] = fn_relative    
+    im['id'] = image_id
+    
+    annotations_this_image = []
+    
+    try:        
+        pil_im = open_image(fn_abs)
+        im_width, im_height = pil_im.size
+        im['width'] = im_width
+        im['height'] = im_height
+        im['error'] = None
+    except Exception as e:
+        print('Warning: error reading {}:\n{}'.format(fn_relative,str(e)))
+        im['width'] = -1
+        im['height'] = -1
+        im['error'] = str(e)
+        return (im,annotations_this_image)
+        
+    # Is there an annotation file for this image?
+    annotation_file = os.path.splitext(fn_abs)[0] + '.txt'
+    if not os.path.isfile(annotation_file):
+        annotation_file = os.path.splitext(fn_abs)[0] + '.TXT'
+    
+    if os.path.isfile(annotation_file):
+        
+        with open(annotation_file,'r') as f:
+            lines = f.readlines()
+        lines = [s.strip() for s in lines]
+        
+        # s = lines[0]
+        annotation_number = 0
+        
+        for s in lines:
+            
+            if len(s.strip()) == 0:
+                continue
+            
+            tokens = s.split()
+            assert len(tokens) == 5
+            category_id = int(tokens[0])
+            assert category_id in category_id_to_name, \
+                'Unrecognized category ID {} in annotation file {}'.format(
+                    category_id,annotation_file)
+            ann = {}
+            ann['id'] = im['id'] + '_' + str(annotation_number)
+            ann['image_id'] = im['id']
+            ann['category_id'] = category_id
+            ann['sequence_level_annotation'] = False
+            
+            # COCO: [x_min, y_min, width, height] in absolute coordinates
+            # YOLO: [class, x_center, y_center, width, height] in normalized coordinates
+            
+            yolo_bbox = [float(x) for x in tokens[1:]]
+            
+            normalized_x_center = yolo_bbox[0]
+            normalized_y_center = yolo_bbox[1]
+            normalized_width = yolo_bbox[2]
+            normalized_height = yolo_bbox[3]
+            
+            absolute_x_center = normalized_x_center * im_width 
+            absolute_y_center = normalized_y_center * im_height
+            absolute_width = normalized_width * im_width
+            absolute_height = normalized_height * im_height
+            absolute_x_min = absolute_x_center - absolute_width / 2
+            absolute_y_min = absolute_y_center - absolute_height / 2
+            
+            coco_bbox = [absolute_x_min, absolute_y_min, absolute_width, absolute_height]
+            
+            ann['bbox'] = coco_bbox
+            annotation_number += 1
+            
+            annotations_this_image.append(ann)                
+            
+        # ...for each annotation 
+        
+    # ...if this image has annotations
+    
+    return (im,annotations_this_image)
+
+# ...def _process_image(...)
+
 
 
 #%% Main conversion function
 
-def yolo_to_coco(input_folder,class_name_file,output_file=None):
+def yolo_to_coco(input_folder,
+                 class_name_file,
+                 output_file=None,
+                 empty_image_handling='no_annotations',
+                 empty_image_category_name='empty',
+                 error_image_handling='no_annotations',
+                 allow_images_without_label_files=True,
+                 n_workers=1,
+                 pool_type='thread',
+                 recursive=True,
+                 exclude_string=None,
+                 include_string=None):
     """
     Convert the YOLO-formatted data in [input_folder] to a COCO-formatted dictionary,
-    reading class names from the flat list [class_name_file].  Optionally writes the output
-    dataset to [output_file].
+    reading class names from [class_name_file], which can be a flat list with a .txt
+    extension or a YOLO dataset.yml file.  Optionally writes the output dataset to [output_file].
+    
+    empty_image_handling can be:
+        
+    * 'no_annotations': include the image in the image list, with no annotations
+    
+    * 'empty_annotations': include the image in the image list, and add an annotation without
+      any bounding boxes, using a category called [empty_image_category_name].
+      
+    * 'skip': don't include the image in the image list
+    
+    * 'error': there shouldn't be any empty images
+    
+    error_image_handling can be:
+        
+    * 'skip': don't include the image at all
+    
+    * 'no_annotations': include with no annotations
+    
+    All images will be assigned an "error" value, usually None.
+        
+    Returns a COCO-formatted dictionary.
     """
     
-    # Validate input
+    ## Validate input
     
     assert os.path.isdir(input_folder)
     assert os.path.isfile(class_name_file)
     
+    assert empty_image_handling in \
+        ('no_annotations','empty_annotations','skip','error'), \
+            'Unrecognized empty image handling spec: {}'.format(empty_image_handling)
+     
+            
+    ## Read class names
     
-    # Class names
-    
-    with open(class_name_file,'r') as f:
-        lines = f.readlines()
-    assert len(lines) > 0, 'Empty class name file {}'.format(class_name_file)
-    lines = [s.strip() for s in lines]
-    assert len(lines[0]) > 0, 'Empty class name file {} (empty first line)'.format(class_name_file)
+    ext = os.path.splitext(class_name_file)[1][1:]
+    assert ext in ('yml','txt','yaml','data'), 'Unrecognized class name file type {}'.format(
+        class_name_file)
     
-    # Blank lines should only appear at the end
-    b_found_blank = False
-    for s in lines:
-        if len(s) == 0:
-            b_found_blank = True
-        elif b_found_blank:
-            raise ValueError('Invalid class name file {}, non-blank line after the last blank line'.format(
-                class_name_file))
-            
-    category_id_to_name = {}
+    if ext in ('txt','data'):
         
-    for i_category_id,category_name in enumerate(lines):
-        assert len(category_name) > 0
-        category_id_to_name[i_category_id] = category_name
+        with open(class_name_file,'r') as f:
+            lines = f.readlines()
+        assert len(lines) > 0, 'Empty class name file {}'.format(class_name_file)
+        class_names = [s.strip() for s in lines]
+        assert len(lines[0]) > 0, 'Empty class name file {} (empty first line)'.format(class_name_file)
         
+        # Blank lines should only appear at the end
+        b_found_blank = False
+        for s in lines:
+            if len(s) == 0:
+                b_found_blank = True
+            elif b_found_blank:
+                raise ValueError('Invalid class name file {}, non-blank line after the last blank line'.format(
+                    class_name_file))
+    
+        category_id_to_name = {}        
+        for i_category_id,category_name in enumerate(class_names):
+            assert len(category_name) > 0
+            category_id_to_name[i_category_id] = category_name
+            
+    else:
         
-    # Enumerate images
+        assert ext in ('yml','yaml')
+        category_id_to_name = read_classes_from_yolo_dataset_file(class_name_file)
     
-    image_files = find_images(input_folder,recursive=False)
+    # Find or create the empty image category, if necessary
+    empty_category_id = None
+    
+    if (empty_image_handling == 'empty_annotations'):
+        category_name_to_id = invert_dictionary(category_id_to_name)
+        if empty_image_category_name in category_name_to_id:
+            empty_category_id = category_name_to_id[empty_image_category_name]
+            print('Using existing empty image category with name {}, ID {}'.format(
+                empty_image_category_name,empty_category_id))            
+        else:
+            empty_category_id = len(category_id_to_name)
+            print('Adding an empty category with name {}, ID {}'.format(
+                empty_image_category_name,empty_category_id))
+            category_id_to_name[empty_category_id] = empty_image_category_name
+            
+            
+    ## Enumerate images
+    
+    print('Enumerating images...')
+    
+    image_files_abs = find_images(input_folder,recursive=recursive,convert_slashes=True)
 
-    images = []
-    annotations = []
+    n_files_original = len(image_files_abs)
+    
+    # Optionally include/exclude images matching specific strings
+    if exclude_string is not None:
+        image_files_abs = [fn for fn in image_files_abs if exclude_string not in fn]
+    if include_string is not None:
+        image_files_abs = [fn for fn in image_files_abs if include_string in fn]
+    
+    if len(image_files_abs) != n_files_original or exclude_string is not None or include_string is not None:
+        n_excluded = n_files_original - len(image_files_abs)
+        print('Excluded {} of {} images based on filenames'.format(n_excluded,n_files_original))
+        
     categories = []
     
     for category_id in category_id_to_name:
@@ -74,79 +255,111 @@ def yolo_to_coco(input_folder,class_name_file,output_file=None):
     info['version'] = '1.0'
     info['description'] = 'Converted from YOLO format'
     
-    # fn = image_files[0]
-    for fn in tqdm(image_files):
-        
-        im = Image.open(fn)
-        im_width, im_height = im.size
-        
-        # Create the image object for this image
-        im = {}
-        fn_relative = os.path.relpath(fn,input_folder)
-        im['file_name'] = fn_relative
-        im['id'] = fn_relative.replace(' ','_')
-        im['location'] = 'unknown'
-        images.append(im)
-        
-        # Is there an annotation file for this image?
-        annotation_file = os.path.splitext(fn)[0] + '.txt'
-        if not os.path.isfile(annotation_file):
-            annotation_file = os.path.splitext(fn)[0] + '.TXT'
-        if not os.path.isfile(annotation_file):
-            # This is an image with no annotations, currently don't do anything special
-            # here
-            pass
+    image_ids = set()
+    
+    
+    ## If we're expected to have labels for every image, check before we process all the images
+    
+    if not allow_images_without_label_files:
+        print('Verifying that label files exist')
+        for image_file_abs in tqdm(image_files_abs):
+            label_file_abs = os.path.splitext(image_file_abs)[0] + '.txt'
+            assert os.path.isfile(label_file_abs), \
+                'No annotation file for {}'.format(image_file_abs)
+    
+    
+    ## Initial loop to make sure image IDs will be unique
+    
+    print('Validating image IDs...')
+    
+    for fn_abs in tqdm(image_files_abs):
+        
+        fn_relative = os.path.relpath(fn_abs,input_folder)
+        image_id = filename_to_image_id(fn_relative)
+        assert image_id not in image_ids, \
+            'Oops, you have hit a very esoteric case where you have the same filename ' + \
+            'with both spaces and underscores, this is not currently handled.'
+        image_ids.add(image_id)
+    
+    
+    ## Main loop to process labels
+    
+    print('Processing labels...')
+    
+    if n_workers <= 1:
+        
+        image_results = []        
+        for fn_abs in tqdm(image_files_abs):                
+            image_results.append(_process_image(fn_abs,input_folder,category_id_to_name))
+            
+    else:
+        
+        assert pool_type in ('process','thread'), 'Illegal pool type {}'.format(pool_type)
+        
+        if pool_type == 'thread':
+            pool = ThreadPool(n_workers)
         else:
-            with open(annotation_file,'r') as f:
-                lines = f.readlines()
-            lines = [s.strip() for s in lines]
-            
-            # s = lines[0]
-            annotation_number = 0
-            for s in lines:
-                if len(s.strip()) == 0:
-                    continue
-                tokens = s.split()
-                assert len(tokens) == 5
-                category_id = int(tokens[0])
-                assert category_id in category_id_to_name, \
-                    'Unrecognized category ID {} in annotation file {}'.format(
-                        category_id,annotation_file)
-                ann = {}
-                ann['id'] = im['id'] + '_' + str(annotation_number)
-                ann['image_id'] = im['id']
-                ann['category_id'] = category_id
-                ann['sequence_level_annotation'] = False
-                
-                # COCO: [x_min, y_min, width, height] in absolute coordinates
-                # YOLO: [class, x_center, y_center, width, height] in normalized coordinates
-                
-                yolo_bbox = [float(x) for x in tokens[1:]]
-                
-                normalized_x_center = yolo_bbox[0]
-                normalized_y_center = yolo_bbox[1]
-                normalized_width = yolo_bbox[2]
-                normalized_height = yolo_bbox[3]
-                
-                absolute_x_center = normalized_x_center * im_width 
-                absolute_y_center = normalized_y_center * im_height
-                absolute_width = normalized_width * im_width
-                absolute_height = normalized_height * im_height
-                absolute_x_min = absolute_x_center - absolute_width / 2
-                absolute_y_min = absolute_y_center - absolute_height / 2
-                
-                coco_bbox = [absolute_x_min, absolute_y_min, absolute_width, absolute_height]
+            pool = Pool(n_workers)
+        
+        print('Starting a {} pool of {} workers'.format(pool_type,n_workers))
+        
+        p = partial(_process_image,input_folder=input_folder,
+                    category_id_to_name=category_id_to_name)
+        image_results = list(tqdm(pool.imap(p, image_files_abs),
+                                  total=len(image_files_abs)))
                 
-                ann['bbox'] = coco_bbox
-                annotation_number += 1
+    
+    assert len(image_results) == len(image_files_abs)
+    
+    
+    ## Re-assembly of results into a COCO dict
+    
+    print('Assembling labels...')
+    
+    images = []
+    annotations = []
+    
+    for image_result in tqdm(image_results):
+    
+        im = image_result[0]
+        annotations_this_image = image_result[1]
+           
+        # If we have annotations for this image
+        if len(annotations_this_image) > 0:
+            assert im['error'] is None
+            images.append(im)
+            for ann in annotations_this_image:
+                annotations.append(ann)
                 
-                annotations.append(ann)                
+        # If this image failed to read
+        elif im['error'] is not None:
+            
+            if error_image_handling == 'skip':
+                pass
+            elif error_image_handling == 'no_annotations':
+                images.append(im)            
                 
-            # ...for each annotation 
+        # If this image read successfully, but there are no annotations
+        else:
             
-        # ...if this image has annotations
+            if empty_image_handling == 'skip':
+                pass
+            elif empty_image_handling == 'no_annotations':
+                images.append(im)
+            elif empty_image_handling == 'empty_annotations':
+                assert empty_category_id  is not None
+                ann = {}
+                ann['id'] = im['id'] + '_0'
+                ann['image_id'] = im['id']
+                ann['category_id'] = empty_category_id
+                ann['sequence_level_annotation'] = False
+                # This would also be a reasonable thing to do, but it's not the convention
+                # we're adopting.
+                # ann['bbox'] = [0,0,0,0]
+                annotations.append(ann)
+                images.append(im)        
         
-    # ...for each image
+    # ...for each image result
     
     print('Read {} annotations for {} images'.format(len(annotations),
                                                      len(images)))

detection/process_video.py

@@ -26,12 +26,17 @@ from detection.video_utils import frame_results_to_video_results
 from detection.video_utils import video_folder_to_frames
 from uuid import uuid1
 
+from detection.video_utils import default_fourcc
+
 
 #%% Options classes
 
 class ProcessVideoOptions:
 
-    model_file = ''
+    # Can be a model filename (.pt or .pb) or a model name (e.g. "MDV5A")
+    model_file = 'MDV5A'
+    
+    # Can be a file or a folder
     input_video_file = ''
 
     output_json_file = None
@@ -72,9 +77,10 @@ class ProcessVideoOptions:
     
     recursive = False 
     verbose = False
+    
     fourcc = None
 
-    rendering_confidence_threshold = 0.15
+    rendering_confidence_threshold = None
     json_confidence_threshold = 0.005
     frame_sample = None
     
@@ -175,8 +181,14 @@ def process_video(options):
             confidence_threshold=options.rendering_confidence_threshold)
 
         # Combine into a video
-        print('Rendering video to {} at {} fps'.format(options.output_video_file,Fs))
-        frames_to_video(detected_frame_files, Fs, options.output_video_file, codec_spec=options.fourcc)
+        if options.frame_sample is None:
+            rendering_fs = Fs
+        else:
+            rendering_fs = Fs / options.frame_sample
+            
+        print('Rendering video to {} at {} fps (original video {} fps)'.format(
+            options.output_video_file,rendering_fs,Fs))
+        frames_to_video(detected_frame_files, rendering_fs, options.output_video_file, codec_spec=options.fourcc)
         
         # Delete the temporary directory we used for detection images
         if not options.keep_rendered_frames:
@@ -344,11 +356,19 @@ def process_video_folder(options):
             output_video_folder = options.input_video_file
                                 
         # For each video
+        #
+        # TODO: parallelize this loop
+        #
         # i_video=0; input_video_file_abs = video_filenames[i_video]
         for i_video,input_video_file_abs in enumerate(video_filenames):
             
             video_fs = Fs[i_video]
             
+            if options.frame_sample is None:                
+                rendering_fs = video_fs
+            else:
+                rendering_fs = video_fs / options.frame_sample
+            
             input_video_file_relative = os.path.relpath(input_video_file_abs,options.input_video_file)
             video_frame_output_folder = os.path.join(frame_rendering_output_dir,input_video_file_relative)
             assert os.path.isdir(video_frame_output_folder), \
@@ -371,11 +391,10 @@ def process_video_folder(options):
             os.makedirs(os.path.dirname(video_output_file),exist_ok=True)
             
             # Create the output video            
-            print('Rendering detections for video {} to {} at {} fps'.format(input_video_file_relative,
-                                                              video_output_file,video_fs))
-            frames_to_video(video_frame_files, video_fs, video_output_file, codec_spec=options.fourcc)
-            
-    
+            print('Rendering detections for video {} to {} at {} fps (original video {} fps)'.format(
+                input_video_file_relative,video_output_file,rendering_fs,video_fs))
+            frames_to_video(video_frame_files, rendering_fs, video_output_file, codec_spec=options.fourcc)
+                
         # ...for each video
         
         # Possibly clean up rendered frames
@@ -525,12 +544,14 @@ if False:
 
 def main():
 
+    default_options = ProcessVideoOptions()
+    
     parser = argparse.ArgumentParser(description=(
         'Run MegaDetector on each frame in a video (or every Nth frame), optionally '\
         'producing a new video with detections annotated'))
 
     parser.add_argument('model_file', type=str,
-                        help='MegaDetector model file')
+                        help='MegaDetector model file (.pt or .pb) or model name (e.g. "MDV5A")')
 
     parser.add_argument('input_video_file', type=str,
                         help='video file (or folder) to process')
@@ -567,8 +588,8 @@ def main():
     parser.add_argument('--render_output_video', action='store_true',
                         help='enable video output rendering (not rendered by default)')
 
-    parser.add_argument('--fourcc', default=None,
-                        help='fourcc code to use for video encoding, only used if render_output_video is True')
+    parser.add_argument('--fourcc', default=default_fourcc,
+                        help='fourcc code to use for video encoding (default {}), only used if render_output_video is True'.format(default_fourcc))
     
     parser.add_argument('--keep_rendered_frames',
                        action='store_true', help='Disable the deletion of rendered (w/boxes) frames')
@@ -586,11 +607,12 @@ def main():
                            'whether other files were present in the folder.')
         
     parser.add_argument('--rendering_confidence_threshold', type=float,
-                        default=0.8, help="don't render boxes with confidence below this threshold")
+                        default=None, help="don't render boxes with confidence below this threshold (defaults to choosing based on the MD version)")
 
     parser.add_argument('--json_confidence_threshold', type=float,
                         default=0.0, help="don't include boxes in the .json file with confidence "\
-                            'below this threshold')
+                            'below this threshold (default {})'.format(
+                                default_options.json_confidence_threshold))
 
     parser.add_argument('--n_cores', type=int,
                         default=1, help='number of cores to use for frame separation and detection. '\

detection/pytorch_detector.py

@@ -234,7 +234,7 @@ class PTDetector:
             if self.device == 'mps':
                 # As of v1.13.0.dev20220824, nms is not implemented for MPS.
                 #
-                # Send predication back to the CPU to fix.
+                # Send prediction back to the CPU to fix.
                 pred = non_max_suppression(prediction=pred.cpu(), conf_thres=detection_threshold)
             else: 
                 pred = non_max_suppression(prediction=pred, conf_thres=detection_threshold)