File size: 8,138 Bytes
3f1bb3e
 
 
 
 
 
bd93c04
3f1bb3e
 
1fb67a9
 
3f1bb3e
a134e44
3f1bb3e
 
 
 
 
 
 
 
 
 
 
 
1fb67a9
 
 
 
 
 
 
 
3f1bb3e
1fb67a9
 
3f1bb3e
 
 
 
1fb67a9
 
 
 
 
 
3f1bb3e
 
 
1fb67a9
 
 
 
 
 
 
 
 
 
 
 
a134e44
 
 
 
1fb67a9
8b5e39e
7b5b426
 
 
8b5e39e
 
 
 
 
 
 
 
 
 
0971307
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8b5e39e
 
7b5b426
 
 
8b5e39e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7b5b426
 
 
8b5e39e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bd93c04
 
 
 
 
 
 
 
 
 
0971307
 
bd93c04
1857544
c0b5959
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
import matplotlib.pyplot as plt
import requests, validators
import torch
import pathlib
import numpy as np
from PIL import Image
import cv2 as cv

from transformers import DetrFeatureExtractor, DetrForSegmentation, MaskFormerImageProcessor, MaskFormerForInstanceSegmentation
# from transformers.models.detr.feature_extraction_detr import rgb_to_id
from transformers.image_transforms import rgb_to_id

TEST_IMAGE = Image.open(r"images/9999999_00783_d_0000358.jpg")
MODEL_NAME_DETR = "facebook/detr-resnet-50-panoptic"
MODEL_NAME_MASKFORMER = "facebook/maskformer-swin-large-coco"
DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")

#######
# Parameters
#######
image = TEST_IMAGE
model_name = MODEL_NAME_MASKFORMER

# Starting with MaskFormer

processor = MaskFormerImageProcessor.from_pretrained(model_name) # <class 'transformers.models.maskformer.image_processing_maskformer.MaskFormerImageProcessor'>
# DIR() --> ['__call__', '__class__', '__delattr__', '__dict__', '__dir__', '__doc__', '__eq__', '__format__', '__ge__', '__getattribute__', '__gt__', '__hash__', '__init__', 
#           '__init_subclass__', '__le__', '__lt__', '__module__', '__ne__', '__new__', '__reduce__', '__reduce_ex__', '__repr__', '__setattr__', '__sizeof__', '__str__', '__subclasshook__', 
#           '__weakref__', '_auto_class', '_create_repo', '_get_files_timestamps', '_max_size', '_pad_image', '_preprocess', '_preprocess_image', '_preprocess_mask', '_processor_class', 
#           '_set_processor_class', '_upload_modified_files', 'center_crop', 'convert_segmentation_map_to_binary_masks', 'do_normalize', 'do_reduce_labels', 'do_rescale', 'do_resize', 
#           'encode_inputs', 'fetch_images', 'from_dict', 'from_json_file', 'from_pretrained', 'get_image_processor_dict', 'ignore_index', 'image_mean', 'image_std', 'model_input_names', 
#           'normalize', 'pad', 'post_process_instance_segmentation', 'post_process_panoptic_segmentation', 'post_process_segmentation', 'post_process_semantic_segmentation', 'preprocess', 
#           'push_to_hub', 'register_for_auto_class', 'resample', 'rescale', 'rescale_factor', 'resize', 'save_pretrained', 'size', 'size_divisor', 'to_dict', 'to_json_file', 'to_json_string']

model = MaskFormerForInstanceSegmentation.from_pretrained(model_name) # <class 'transformers.models.maskformer.modeling_maskformer.MaskFormerForInstanceSegmentation'>
# DIR for model was too big
model.to(DEVICE)

# img = np.array(TEST_IMAGE)

inputs = processor(images=image, return_tensors="pt") # <class 'transformers.image_processing_utils.BatchFeature'>
# DIR() --> ['_MutableMapping__marker', '__abstractmethods__', '__class__', '__contains__', '__copy__', '__delattr__', '__delitem__', '__dict__', '__dir__', '__doc__', '__eq__', '__format__', 
#           '__ge__', '__getattr__', '__getattribute__', '__getitem__', '__getstate__', '__gt__', '__hash__', '__init__', '__init_subclass__', '__iter__', '__le__', '__len__', '__lt__', 
#           '__module__', '__ne__', '__new__', '__reduce__', '__reduce_ex__', '__repr__', '__reversed__', '__setattr__', '__setitem__', '__setstate__', '__sizeof__', '__slots__', '__str__', 
#           '__subclasshook__', '__weakref__', '_abc_impl', '_get_is_as_tensor_fns', 'clear', 'convert_to_tensors', 'copy', 'data', 'fromkeys', 'get', 'items', 'keys', 'pop', 'popitem', 
#           'setdefault', 'to', 'update', 'values']
inputs.to(DEVICE)


outputs = model(**inputs) # <class 'transformers.models.maskformer.modeling_maskformer.MaskFormerForInstanceSegmentationOutput'>
# Each element of this class is a <class 'torch.Tensor'>
# DIR() --> ['__annotations__', '__class__', '__contains__', '__dataclass_fields__', '__dataclass_params__', '__delattr__', '__delitem__', '__dict__', '__dir__', 
#           '__doc__', '__eq__', '__format__', '__ge__', '__getattribute__', '__getitem__', '__gt__', '__hash__', '__init__', '__init_subclass__', '__iter__', 
#           '__le__', '__len__', '__lt__', '__module__', '__ne__', '__new__', '__post_init__', '__reduce__', '__reduce_ex__', '__repr__', '__reversed__', '__setattr__', 
#           '__setitem__', '__sizeof__', '__str__', '__subclasshook__', 'attentions', 'auxiliary_logits', 'class_queries_logits', 'clear', 'copy', 'encoder_hidden_states', 
#           'encoder_last_hidden_state', 'fromkeys', 'get', 'hidden_states', 'items', 'keys', 'loss', 'masks_queries_logits', 'move_to_end', 'pixel_decoder_hidden_states', 
#           'pixel_decoder_last_hidden_state', 'pop', 'popitem', 'setdefault', 'to_tuple', 'transformer_decoder_hidden_states', 'transformer_decoder_last_hidden_state', 
#           'update', 'values']

results = processor.post_process_panoptic_segmentation(outputs, target_sizes=[image.size[::-1]])[0]
# <class 'dict'>
# Keys: dict_keys(['segmentation', 'segments_info'])
# type(results["segments_info"]) --> list
# type(results["segmentation"]) --> <class 'torch.Tensor'>


def show_mask_for_number(map_to_use, label_id):
    """
    map_to_use: You have to pass in `results["segmentation"]`
    """
    if torch.cuda.is_available():
        mask = (map_to_use.cpu().numpy() == label_id)
    else:
        mask = (map_to_use.numpy() == label_id)
    
    visual_mask = (mask* 255).astype(np.uint8)
    visual_mask = Image.fromarray(visual_mask)
    plt.imshow(visual_mask)
    plt.show()

def show_mask_for_number_over_image(map_to_use, label_id, image_object):
    """
    map_to_use: You have to pass in `results["segmentation"]`
    """
    if torch.cuda.is_available():
        mask = (map_to_use.cpu().numpy() == label_id)
    else:
        mask = (map_to_use.numpy() == label_id)
    
    visual_mask = (mask* 255).astype(np.uint8)
    visual_mask = Image.fromarray(visual_mask)
    plt.imshow(image_object)
    plt.imshow(visual_mask, alpha=0.25)
    plt.show()


def get_coordinates_for_bb_simple(map_to_use, label_id):
    """
    map_to_use: You have to pass in `results["segmentation"]`
    """
    if torch.cuda.is_available():
        mask = (map_to_use.cpu().numpy() == label_id)
    else:
        mask = (map_to_use.numpy() == label_id)
    
    x, y = np.where(mask==True)
    x_max, x_min = max(x), min(x)
    y_max, y_min = max(y), min(y)
    return (x_min, y_min), (x_max, y_max)

def make_simple_box(left_top, right_bottom, map_size):
    full_mask = np.full(map_size, False)
    left_x, top_y = left_top
    right_x, bottom_y = right_bottom
    full_mask[left_x:right_x, top_y] = True
    full_mask[left_x:right_x, bottom_y] = True
    full_mask[left_x, top_y:bottom_y] = True
    full_mask[right_x, top_y:bottom_y] = True

    visual_mask = (full_mask* 255).astype(np.uint8)
    visual_mask = Image.fromarray(visual_mask)
    plt.imshow(visual_mask)
    plt.show()


def test(map_to_use, label_id):
    """
    map_to_use: You have to pass in `results["segmentation"]`
    """
    if torch.cuda.is_available():
        mask = (map_to_use.cpu().numpy() == label_id)
    else:
        mask = (map_to_use.numpy() == label_id)
    
    
    lt, rb = get_coordinates_for_bb_simple(map_to_use, label_id)
    left_x, top_y = lt
    right_x, bottom_y = rb
    
    mask[left_x:right_x, top_y] = .5
    mask[left_x:right_x, bottom_y] = .5
    mask[left_x, top_y:bottom_y] = .5
    mask[right_x, top_y:bottom_y] = .5

    visual_mask = (mask* 255).astype(np.uint8)
    visual_mask = Image.fromarray(visual_mask)
    plt.imshow(visual_mask)
    plt.show()

def contour_map(map_to_use, label_id):
    """
    map_to_use: You have to pass in `results["segmentation"]`
    """
    if torch.cuda.is_available():
        mask = (map_to_use.cpu().numpy() == label_id)
    else:
        mask = (map_to_use.numpy() == label_id)
    
    visual_mask = (mask* 255).astype(np.uint8)
    contours, hierarchy = cv.findContours(visual_mask, cv.RETR_TREE, cv.CHAIN_APPROX_SIMPLE)
    return contours, hierarchy


# https://docs.opencv.org/4.9.0/dd/d49/tutorial_py_contour_features.html


# Idea for determining if close
#   https://dsp.stackexchange.com/questions/2564/opencv-c-connect-nearby-contours-based-on-distance-between-them
#   Bing Search: cv determine if 2 contours belong together