Skip to content

Coco assistant

coco_assistant.coco_assistant

COCO_Assistant

COCO_Assistant object

Source code in coco_assistant/coco_assistant.py 
 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
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
class COCO_Assistant:
    """COCO_Assistant object"""

    def __init__(self, img_dir, ann_dir):
        """

        Args:
            img_dir (str): Path to images folder.
            ann_dir (str): Path to images folder.

        """
        self.img_dir = Path(img_dir)
        self.ann_dir = Path(ann_dir)
        self.res_dir = self.ann_dir.parent / "results"

        self.dh = utils.DirectoryHandler(img_dir, ann_dir, self.res_dir)

        # TODO: Add check for confirming these folders only contain .jpg and .json respectively
        logging.debug("Number of image folders = %s", len(self.dh.names))
        logging.debug("Number of annotation files = %s", len(self.dh.names))

        self.annfiles = [COCO(self.ann_dir / (i + ".json")) for i in self.dh.names]
        self.anndict = dict(zip(self.dh.names, self.annfiles))

        self.ann_anchors = []

    def merge(self):
        """
        Merge multiple coco datasets
        """

        resann_dir = self.dh.create("merged/annotations")

        cann = {"images": [], "annotations": [], "info": None, "licenses": None, "categories": None}

        logging.debug("Merging Annotations...")

        dst_ann = resann_dir / "merged.json"

        print("Merging annotations")
        for j in tqdm(self.dh.names):
            cj = self.anndict[j].dataset

            ind = self.dh.names.index(j)
            # Check if this is the 1st annotation.
            # If it is, continue else modify current annotation
            if ind == 0:
                cann["images"] = cann["images"] + cj["images"]
                cann["annotations"] = cann["annotations"] + cj["annotations"]
                if "info" in list(cj.keys()):
                    cann["info"] = cj["info"]
                if "licenses" in list(cj.keys()):
                    cann["licenses"] = cj["licenses"]
                cann["categories"] = sorted(cj["categories"], key=lambda i: i["id"])

                last_imid = cann["images"][-1]["id"]
                last_annid = cann["annotations"][-1]["id"]

                # If last imid or last_annid is a str, convert it to int
                if isinstance(last_imid, str) or isinstance(last_annid, str):
                    logging.debug("String Ids detected. Converting to int")
                    id_dict = {}
                    # Change image id in images field
                    for i, im in enumerate(cann["images"]):
                        id_dict[im["id"]] = i
                        im["id"] = i

                    # Change annotation id & image id in annotations field
                    for i, im in enumerate(cann["annotations"]):
                        im["id"] = i
                        if isinstance(last_imid, str):
                            im["image_id"] = id_dict[im["image_id"]]

                last_imid = max(im["id"] for im in cann["images"])
                last_annid = max(ann["id"] for ann in cann["annotations"])

            else:

                id_dict = {}
                # Change image id in images field
                for i, im in enumerate(cj["images"]):
                    id_dict[im["id"]] = last_imid + i + 1
                    im["id"] = last_imid + i + 1

                # Change annotation and image ids in annotations field
                for i, ann in enumerate(cj["annotations"]):
                    ann["id"] = last_annid + i + 1
                    ann["image_id"] = id_dict[ann["image_id"]]

                # Remap categories
                cmapper = utils.CatRemapper(cann["categories"], cj["categories"])
                cann["categories"], cj["annotations"] = cmapper.remap(cj["annotations"])

                cann["images"] = cann["images"] + cj["images"]
                cann["annotations"] = cann["annotations"] + cj["annotations"]
                if "info" in list(cj.keys()):
                    cann["info"] = cj["info"]
                if "licenses" in list(cj.keys()):
                    cann["licenses"] = cj["licenses"]

                last_imid = cann["images"][-1]["id"]
                last_annid = cann["annotations"][-1]["id"]

        with open(dst_ann, "w") as aw:
            json.dump(cann, aw)

    def remove_cat(self, interactive=True, jc=None, rcats=None):

        """
        Remove categories.

        In interactive mode, you can input the json and the categories to be
        removed (as a list, see Usage for example)

        In non-interactive mode, you manually pass in json filename and
        categories to be removed. Note that jc and rcats cannot be None if run
        with interactive=False.

        Raises:
            AssertionError: if specified index exceeds number of datasets
            AssertionError: if rcats is not a list of strings
            AssertionError: if jc = rcats = None
        """

        resrm_dir = self.dh.create("removal")

        if interactive:
            print(self.dh.names)
            print("Choose directory index (1:first, 2: second ..)")

            json_choice = input()
            try:
                json_choice = int(json_choice) - 1
            except ValueError:
                sys.exit("Please specify an index")

            if json_choice > len(self.dh.names):
                raise AssertionError("Index exceeds number of datasets")
            # ann = self.annfiles[json_choice]
            name = self.dh.names[json_choice]
            json_name = name + ".json"
            ann = self.anndict[name]

            print("\nCategories present:")
            cats = [i["name"] for i in ann.cats.values()]
            print(cats)

            self.rcats = []
            print("\nEnter categories you wish to remove as a list:")
            x = input()
            x = ast.literal_eval(x)
            if isinstance(x, list) is False:
                raise AssertionError("Input must be a list of categories to be removed")
            if all(elem in cats for elem in x):
                self.rcats = x
            else:
                print("Incorrect entry.")

        else:
            if jc is None or rcats is None:
                raise AssertionError(
                    "Both json choice and rcats need to be provided in non-interactive mode"
                )
            # If passed, json_choice needs to be full path
            json_choice = Path(jc)  # Full path
            name = json_choice.stem
            json_name = json_choice.name  # abc
            ann = self.anndict[name]
            self.rcats = rcats

        print("Removing specified categories...")

        # Gives you a list of category ids of the categories to be removed
        catids_remove = ann.getCatIds(catNms=self.rcats)
        # Gives you a list of ids of annotations that contain those categories
        annids_remove = ann.getAnnIds(catIds=catids_remove)

        # Get keep category ids
        catids_keep = list(set(ann.getCatIds()) - set(catids_remove))
        # Get keep annotation ids
        annids_keep = list(set(ann.getAnnIds()) - set(annids_remove))

        with open(self.ann_dir / json_name) as it:
            x = json.load(it)

        del x["categories"]
        x["categories"] = ann.loadCats(catids_keep)
        del x["annotations"]
        x["annotations"] = ann.loadAnns(annids_keep)

        with open(resrm_dir / json_name, "w") as oa:
            json.dump(x, oa)

    def ann_stats(self, stat, arearng, show_count=False, save=False):
        """Display statistics.

        Args:
            stat (str): Type of statistic to be shown. Supports ["area", "cat"]
            arearng (list[float]): Area range list
            show_count (bool, optional): Shows category countplot if True. Defaults to False.
            save (bool, optional): Save stat plot to disk if True. Defaults to False.
        """
        if stat == "area":
            stats.pi_area_split(self.anndict, areaRng=arearng, save=save)
        elif stat == "cat":
            stats.cat_count(self.anndict, show_count=show_count, save=save)

    def anchors(self, n, fmt="rect", recompute=False):
        """
        Generate top N anchors

        !!! attention
            Experimental feature

        Args:
            n (int): Number of anchors
            fmt (str): Anchor type i.e. square or rectangular. Defaults to "rect".
            recompute (bool, optional): Recomputes the anchors if True. Defaults to False.
        """

        if recompute or not self.ann_anchors:
            print("Calculating anchors...")
            names, anns = self.anndict.keys(), self.anndict.values()
            a = [utils.generate_anchors(j, n, fmt) for j in anns]
            self.ann_anchors = dict(zip(names, a))
        else:
            print("Loading pre-computed anchors")
            print(self.ann_anchors)

    def get_segmasks(self, palette=True):
        """
        Generate segmentation masks

        Args:
            palette (bool, optional): Create masks with color palette if True. Defaults to True.
        """
        for name, ann in self.anndict.items():
            output_dir = self.res_dir / "segmasks" / name
            utils.det2seg(ann, output_dir, palette)

    def visualise(self):
        """
        Visualise annotations.
        """
        print("Choose directory index (1:first, 2: second ..):")
        print(self.dh.names)

        dir_choice = input()
        try:
            dir_choice = int(dir_choice) - 1
        except ValueError:
            sys.exit("Please specify an index")

        if dir_choice > len(self.dh.names):
            raise AssertionError("Index exceeds number of datasets")
        dir_choice = self.dh.names[dir_choice]
        ann = self.anndict[dir_choice]
        img_dir = self.img_dir / dir_choice
        cocovis.visualise_all(ann, img_dir)

__init__(img_dir, ann_dir)

Parameters:

Name Type Description Default
img_dir str

Path to images folder.

 required
ann_dir str

Path to images folder.

 required
Source code in coco_assistant/coco_assistant.py 
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
def __init__(self, img_dir, ann_dir):
    """

    Args:
        img_dir (str): Path to images folder.
        ann_dir (str): Path to images folder.

    """
    self.img_dir = Path(img_dir)
    self.ann_dir = Path(ann_dir)
    self.res_dir = self.ann_dir.parent / "results"

    self.dh = utils.DirectoryHandler(img_dir, ann_dir, self.res_dir)

    # TODO: Add check for confirming these folders only contain .jpg and .json respectively
    logging.debug("Number of image folders = %s", len(self.dh.names))
    logging.debug("Number of annotation files = %s", len(self.dh.names))

    self.annfiles = [COCO(self.ann_dir / (i + ".json")) for i in self.dh.names]
    self.anndict = dict(zip(self.dh.names, self.annfiles))

    self.ann_anchors = []

anchors(n, fmt='rect', recompute=False)

Generate top N anchors

Attention

Experimental feature

Parameters:

Name Type Description Default
n int

Number of anchors

 required
fmt str

Anchor type i.e. square or rectangular. Defaults to "rect".

 'rect'
recompute bool

Recomputes the anchors if True. Defaults to False.

 False
Source code in coco_assistant/coco_assistant.py 
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
def anchors(self, n, fmt="rect", recompute=False):
    """
    Generate top N anchors

    !!! attention
        Experimental feature

    Args:
        n (int): Number of anchors
        fmt (str): Anchor type i.e. square or rectangular. Defaults to "rect".
        recompute (bool, optional): Recomputes the anchors if True. Defaults to False.
    """

    if recompute or not self.ann_anchors:
        print("Calculating anchors...")
        names, anns = self.anndict.keys(), self.anndict.values()
        a = [utils.generate_anchors(j, n, fmt) for j in anns]
        self.ann_anchors = dict(zip(names, a))
    else:
        print("Loading pre-computed anchors")
        print(self.ann_anchors)

ann_stats(stat, arearng, show_count=False, save=False)

Display statistics.

Parameters:

Name Type Description Default
stat str

Type of statistic to be shown. Supports ["area", "cat"]

 required
arearng list[float]

Area range list

 required
show_count bool

Shows category countplot if True. Defaults to False.

 False
save bool

Save stat plot to disk if True. Defaults to False.

 False
Source code in coco_assistant/coco_assistant.py 
231
232
233
234
235
236
237
238
239
240
241
242
243
def ann_stats(self, stat, arearng, show_count=False, save=False):
    """Display statistics.

    Args:
        stat (str): Type of statistic to be shown. Supports ["area", "cat"]
        arearng (list[float]): Area range list
        show_count (bool, optional): Shows category countplot if True. Defaults to False.
        save (bool, optional): Save stat plot to disk if True. Defaults to False.
    """
    if stat == "area":
        stats.pi_area_split(self.anndict, areaRng=arearng, save=save)
    elif stat == "cat":
        stats.cat_count(self.anndict, show_count=show_count, save=save)

get_segmasks(palette=True)

Generate segmentation masks

Parameters:

Name Type Description Default
palette bool

Create masks with color palette if True. Defaults to True.

 True
Source code in coco_assistant/coco_assistant.py 
267
268
269
270
271
272
273
274
275
276
def get_segmasks(self, palette=True):
    """
    Generate segmentation masks

    Args:
        palette (bool, optional): Create masks with color palette if True. Defaults to True.
    """
    for name, ann in self.anndict.items():
        output_dir = self.res_dir / "segmasks" / name
        utils.det2seg(ann, output_dir, palette)

merge()

Merge multiple coco datasets

Source code in coco_assistant/coco_assistant.py 
 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
def merge(self):
    """
    Merge multiple coco datasets
    """

    resann_dir = self.dh.create("merged/annotations")

    cann = {"images": [], "annotations": [], "info": None, "licenses": None, "categories": None}

    logging.debug("Merging Annotations...")

    dst_ann = resann_dir / "merged.json"

    print("Merging annotations")
    for j in tqdm(self.dh.names):
        cj = self.anndict[j].dataset

        ind = self.dh.names.index(j)
        # Check if this is the 1st annotation.
        # If it is, continue else modify current annotation
        if ind == 0:
            cann["images"] = cann["images"] + cj["images"]
            cann["annotations"] = cann["annotations"] + cj["annotations"]
            if "info" in list(cj.keys()):
                cann["info"] = cj["info"]
            if "licenses" in list(cj.keys()):
                cann["licenses"] = cj["licenses"]
            cann["categories"] = sorted(cj["categories"], key=lambda i: i["id"])

            last_imid = cann["images"][-1]["id"]
            last_annid = cann["annotations"][-1]["id"]

            # If last imid or last_annid is a str, convert it to int
            if isinstance(last_imid, str) or isinstance(last_annid, str):
                logging.debug("String Ids detected. Converting to int")
                id_dict = {}
                # Change image id in images field
                for i, im in enumerate(cann["images"]):
                    id_dict[im["id"]] = i
                    im["id"] = i

                # Change annotation id & image id in annotations field
                for i, im in enumerate(cann["annotations"]):
                    im["id"] = i
                    if isinstance(last_imid, str):
                        im["image_id"] = id_dict[im["image_id"]]

            last_imid = max(im["id"] for im in cann["images"])
            last_annid = max(ann["id"] for ann in cann["annotations"])

        else:

            id_dict = {}
            # Change image id in images field
            for i, im in enumerate(cj["images"]):
                id_dict[im["id"]] = last_imid + i + 1
                im["id"] = last_imid + i + 1

            # Change annotation and image ids in annotations field
            for i, ann in enumerate(cj["annotations"]):
                ann["id"] = last_annid + i + 1
                ann["image_id"] = id_dict[ann["image_id"]]

            # Remap categories
            cmapper = utils.CatRemapper(cann["categories"], cj["categories"])
            cann["categories"], cj["annotations"] = cmapper.remap(cj["annotations"])

            cann["images"] = cann["images"] + cj["images"]
            cann["annotations"] = cann["annotations"] + cj["annotations"]
            if "info" in list(cj.keys()):
                cann["info"] = cj["info"]
            if "licenses" in list(cj.keys()):
                cann["licenses"] = cj["licenses"]

            last_imid = cann["images"][-1]["id"]
            last_annid = cann["annotations"][-1]["id"]

    with open(dst_ann, "w") as aw:
        json.dump(cann, aw)

remove_cat(interactive=True, jc=None, rcats=None)

Remove categories.

In interactive mode, you can input the json and the categories to be removed (as a list, see Usage for example)

In non-interactive mode, you manually pass in json filename and categories to be removed. Note that jc and rcats cannot be None if run with interactive=False.

Raises:

Type Description
AssertionError

if specified index exceeds number of datasets
AssertionError

if rcats is not a list of strings
AssertionError

if jc = rcats = None
Source code in coco_assistant/coco_assistant.py 
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
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
def remove_cat(self, interactive=True, jc=None, rcats=None):

    """
    Remove categories.

    In interactive mode, you can input the json and the categories to be
    removed (as a list, see Usage for example)

    In non-interactive mode, you manually pass in json filename and
    categories to be removed. Note that jc and rcats cannot be None if run
    with interactive=False.

    Raises:
        AssertionError: if specified index exceeds number of datasets
        AssertionError: if rcats is not a list of strings
        AssertionError: if jc = rcats = None
    """

    resrm_dir = self.dh.create("removal")

    if interactive:
        print(self.dh.names)
        print("Choose directory index (1:first, 2: second ..)")

        json_choice = input()
        try:
            json_choice = int(json_choice) - 1
        except ValueError:
            sys.exit("Please specify an index")

        if json_choice > len(self.dh.names):
            raise AssertionError("Index exceeds number of datasets")
        # ann = self.annfiles[json_choice]
        name = self.dh.names[json_choice]
        json_name = name + ".json"
        ann = self.anndict[name]

        print("\nCategories present:")
        cats = [i["name"] for i in ann.cats.values()]
        print(cats)

        self.rcats = []
        print("\nEnter categories you wish to remove as a list:")
        x = input()
        x = ast.literal_eval(x)
        if isinstance(x, list) is False:
            raise AssertionError("Input must be a list of categories to be removed")
        if all(elem in cats for elem in x):
            self.rcats = x
        else:
            print("Incorrect entry.")

    else:
        if jc is None or rcats is None:
            raise AssertionError(
                "Both json choice and rcats need to be provided in non-interactive mode"
            )
        # If passed, json_choice needs to be full path
        json_choice = Path(jc)  # Full path
        name = json_choice.stem
        json_name = json_choice.name  # abc
        ann = self.anndict[name]
        self.rcats = rcats

    print("Removing specified categories...")

    # Gives you a list of category ids of the categories to be removed
    catids_remove = ann.getCatIds(catNms=self.rcats)
    # Gives you a list of ids of annotations that contain those categories
    annids_remove = ann.getAnnIds(catIds=catids_remove)

    # Get keep category ids
    catids_keep = list(set(ann.getCatIds()) - set(catids_remove))
    # Get keep annotation ids
    annids_keep = list(set(ann.getAnnIds()) - set(annids_remove))

    with open(self.ann_dir / json_name) as it:
        x = json.load(it)

    del x["categories"]
    x["categories"] = ann.loadCats(catids_keep)
    del x["annotations"]
    x["annotations"] = ann.loadAnns(annids_keep)

    with open(resrm_dir / json_name, "w") as oa:
        json.dump(x, oa)

visualise()

Visualise annotations.

Source code in coco_assistant/coco_assistant.py 
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
def visualise(self):
    """
    Visualise annotations.
    """
    print("Choose directory index (1:first, 2: second ..):")
    print(self.dh.names)

    dir_choice = input()
    try:
        dir_choice = int(dir_choice) - 1
    except ValueError:
        sys.exit("Please specify an index")

    if dir_choice > len(self.dh.names):
        raise AssertionError("Index exceeds number of datasets")
    dir_choice = self.dh.names[dir_choice]
    ann = self.anndict[dir_choice]
    img_dir = self.img_dir / dir_choice
    cocovis.visualise_all(ann, img_dir)