Adding coco functionality (#523)
This commit is contained in:
PatrickBue
GitHub
Родитель
735f7ff474
Коммит
198b985581
|
|
@ -36,6 +36,7 @@ dependencies:
|
|||
- pre-commit>=1.14.4
|
||||
- pyyaml>=5.1.2
|
||||
- requests>=2.22.0
|
||||
- cytoolz
|
||||
- pip:
|
||||
- nvidia-ml-py3
|
||||
- nteract-scrapbook
|
||||
|
|
|
|||
|
|
@ -239,6 +239,14 @@
|
|||
"You'll notice that inside the annotation xml file, we can see which image the file references `<path>`, the number of `<objects>` in the image, that the image is of (`<name>`) and the bounding box of that object (`<bndbox>`)."
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"#### COCO format \n",
|
||||
"A second popular annotation standard uses the COCO format introduced by the [COCO](http://cocodataset.org) challenge. Annotations in this format need to be converted first to Pascal VOC syntax in order to run this notebook. The function `coco2voc` does exactly that. See the notebook [04_coco_accuracy_vs_speed.ipynb](04_coco_accuracy_vs_speed.ipynb) for a code example."
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
|
|
|
|||
Различия файлов скрыты, потому что одна или несколько строк слишком длинны
|
|
@ -20,6 +20,8 @@ from fastai.vision import cnn_learner, DatasetType, models
|
|||
from fastai.vision.data import ImageList, imagenet_stats
|
||||
from typing import List, Tuple
|
||||
from tempfile import TemporaryDirectory
|
||||
|
||||
from .resources import coco_sample
|
||||
from utils_cv.common.data import unzip_url
|
||||
from utils_cv.common.gpu import db_num_workers
|
||||
from utils_cv.classification.data import Urls as ic_urls
|
||||
|
|
@ -148,6 +150,7 @@ def detection_notebooks():
|
|||
"01": os.path.join(folder_notebooks, "01_training_introduction.ipynb"),
|
||||
"02": os.path.join(folder_notebooks, "02_mask_rcnn.ipynb"),
|
||||
"03": os.path.join(folder_notebooks, "03_keypoint_rcnn.ipynb"),
|
||||
"04": os.path.join(folder_notebooks, "04_coco_accuracy_vs_speed.ipynb"),
|
||||
"11": os.path.join(
|
||||
folder_notebooks, "11_exploring_hyperparameters_on_azureml.ipynb"
|
||||
),
|
||||
|
|
@ -684,6 +687,14 @@ def od_detections(od_detection_dataset):
|
|||
return learner.predict_dl(od_detection_dataset.test_dl, threshold=0)
|
||||
|
||||
|
||||
@pytest.fixture(scope="session")
|
||||
def coco_sample_path(tmpdir_factory) -> str:
|
||||
""" Returns the path to a coco-formatted annotation. """
|
||||
path = tmpdir_factory.mktemp("data").join("coco_sample.json")
|
||||
path.write_text(coco_sample, encoding=None)
|
||||
return path
|
||||
|
||||
|
||||
# ----- AML Settings ----------------------------------------------------------
|
||||
|
||||
|
||||
|
|
|
|||
|
|
@ -0,0 +1,131 @@
|
|||
# Copyright (c) Microsoft Corporation. All rights reserved.
|
||||
# Licensed under the MIT License.
|
||||
|
||||
# NOTE: This file is used by pytest to inject fixtures automatically. As it is explained in the documentation
|
||||
# https://docs.pytest.org/en/latest/fixture.html:
|
||||
# "If during implementing your tests you realize that you want to use a fixture function from multiple test files
|
||||
# you can move it to a conftest.py file. You don't need to import the fixture you want to use in a test, it
|
||||
# automatically gets discovered by pytest."
|
||||
|
||||
coco_sample = """
|
||||
{
|
||||
"info": {
|
||||
"year": "2020",
|
||||
"version": "1",
|
||||
"description": "",
|
||||
"contributor": "",
|
||||
"url": "https://url",
|
||||
"date_created": "2020-01-01T00:00:00+00:00"
|
||||
},
|
||||
"licenses": [
|
||||
{
|
||||
"id": 1,
|
||||
"url": "",
|
||||
"name": "name"
|
||||
}
|
||||
],
|
||||
"categories": [
|
||||
{
|
||||
"id": 0,
|
||||
"name": "bottle",
|
||||
"supercategory": "none"
|
||||
},
|
||||
{
|
||||
"id": 1,
|
||||
"name": "can",
|
||||
"supercategory": "cells"
|
||||
},
|
||||
{
|
||||
"id": 2,
|
||||
"name": "carton",
|
||||
"supercategory": "cells"
|
||||
}
|
||||
],
|
||||
"images": [
|
||||
{
|
||||
"id": 0,
|
||||
"license": 1,
|
||||
"file_name": "image0.jpg",
|
||||
"height": 480,
|
||||
"width": 640,
|
||||
"date_captured": "2020-01-01T00:00:00+00:00",
|
||||
"coco_url": "http://image0.jpg"
|
||||
},
|
||||
{
|
||||
"id": 1,
|
||||
"license": 1,
|
||||
"file_name": "image1.jpg",
|
||||
"height": 480,
|
||||
"width": 640,
|
||||
"date_captured": "2020-01-01T00:00:00+00:00",
|
||||
"url": "http://image1.jpg"
|
||||
},
|
||||
{
|
||||
"id": 2,
|
||||
"license": 1,
|
||||
"file_name": "image2.jpg",
|
||||
"height": 480,
|
||||
"width": 640,
|
||||
"date_captured": "2020-01-01T00:00:00+00:00"
|
||||
}
|
||||
],
|
||||
"annotations": [
|
||||
{
|
||||
"id": 0,
|
||||
"image_id": 0,
|
||||
"category_id": 2,
|
||||
"bbox": [
|
||||
100,
|
||||
200,
|
||||
300,
|
||||
400
|
||||
],
|
||||
"area": 10000,
|
||||
"segmentation": [],
|
||||
"iscrowd": 0
|
||||
},
|
||||
{
|
||||
"id": 1,
|
||||
"image_id": 0,
|
||||
"category_id": 1,
|
||||
"bbox": [
|
||||
100,
|
||||
200,
|
||||
300,
|
||||
400
|
||||
],
|
||||
"area": 10000,
|
||||
"segmentation": [],
|
||||
"iscrowd": 0
|
||||
},
|
||||
{
|
||||
"id": 2,
|
||||
"image_id": 1,
|
||||
"category_id": 1,
|
||||
"bbox": [
|
||||
100,
|
||||
200,
|
||||
300,
|
||||
400
|
||||
],
|
||||
"area": 10000,
|
||||
"segmentation": [],
|
||||
"iscrowd": 0
|
||||
},
|
||||
{
|
||||
"id": 3,
|
||||
"image_id": 2,
|
||||
"category_id": 1,
|
||||
"bbox": [
|
||||
100,
|
||||
200,
|
||||
300,
|
||||
400
|
||||
],
|
||||
"area": 10000,
|
||||
"segmentation": [],
|
||||
"iscrowd": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
"""
|
||||
|
|
@ -3,6 +3,7 @@
|
|||
|
||||
import hashlib
|
||||
import numpy as np
|
||||
import os
|
||||
import pytest
|
||||
import requests
|
||||
|
||||
|
|
@ -12,12 +13,50 @@ import xml.etree.ElementTree as ET
|
|||
|
||||
from utils_cv.detection.data import (
|
||||
coco_labels,
|
||||
coco2voc,
|
||||
Urls,
|
||||
extract_keypoints_from_labelbox_json,
|
||||
extract_masks_from_labelbox_json,
|
||||
)
|
||||
|
||||
|
||||
def test_urls():
|
||||
# Test if all urls are valid
|
||||
all_urls = Urls.all()
|
||||
for url in all_urls:
|
||||
with requests.get(url):
|
||||
pass
|
||||
|
||||
|
||||
def test_coco_labels():
|
||||
# Compare first five labels for quick check
|
||||
COCO_LABELS_FIRST_FIVE = (
|
||||
"__background__",
|
||||
"person",
|
||||
"bicycle",
|
||||
"car",
|
||||
"motorcycle",
|
||||
)
|
||||
|
||||
labels = coco_labels()
|
||||
for i in range(5):
|
||||
assert labels[i] == COCO_LABELS_FIRST_FIVE[i]
|
||||
|
||||
# Check total number of labels
|
||||
assert len(labels) == 91
|
||||
|
||||
|
||||
def test_coco2voc(coco_sample_path):
|
||||
output_dir = "coco2voc_output"
|
||||
coco2voc(
|
||||
anno_path = coco_sample_path,
|
||||
output_dir = output_dir,
|
||||
download_images = False
|
||||
)
|
||||
filenames = os.listdir(os.path.join(output_dir, "annotations"))
|
||||
assert len(filenames) == 3
|
||||
|
||||
|
||||
@pytest.fixture(scope="session")
|
||||
def labelbox_export_data(tmp_session):
|
||||
tmp_session = Path(tmp_session)
|
||||
|
|
@ -146,32 +185,6 @@ def labelbox_export_data(tmp_session):
|
|||
return data_dir, mask_json_path, keypoint_json_path, keypoint_truth_dict
|
||||
|
||||
|
||||
def test_urls():
|
||||
# Test if all urls are valid
|
||||
all_urls = Urls.all()
|
||||
for url in all_urls:
|
||||
with requests.get(url):
|
||||
pass
|
||||
|
||||
|
||||
def test_coco_labels():
|
||||
# Compare first five labels for quick check
|
||||
COCO_LABELS_FIRST_FIVE = (
|
||||
"__background__",
|
||||
"person",
|
||||
"bicycle",
|
||||
"car",
|
||||
"motorcycle",
|
||||
)
|
||||
|
||||
labels = coco_labels()
|
||||
for i in range(5):
|
||||
assert labels[i] == COCO_LABELS_FIRST_FIVE[i]
|
||||
|
||||
# Check total number of labels
|
||||
assert len(labels) == 91
|
||||
|
||||
|
||||
def test_extract_keypoints_from_labelbox_json(
|
||||
labelbox_export_data, tmp_session
|
||||
):
|
||||
|
|
|
|||
|
|
@ -112,6 +112,26 @@ def test_03_notebook_run(
|
|||
)
|
||||
|
||||
|
||||
@pytest.mark.gpu
|
||||
@pytest.mark.notebooks
|
||||
def test_04_notebook_run(detection_notebooks, tiny_od_data_path):
|
||||
notebook_path = detection_notebooks["04"]
|
||||
pm.execute_notebook(
|
||||
notebook_path,
|
||||
OUTPUT_NOTEBOOK,
|
||||
parameters=dict(
|
||||
PM_VERSION=pm.__version__,
|
||||
DATA_PATH=tiny_od_data_path,
|
||||
LABELS=["can", "carton", "milk_bottle", "water_bottle"]*21 #coco model was pre-trained on 80 classes
|
||||
),
|
||||
kernel_name=KERNEL_NAME,
|
||||
)
|
||||
|
||||
nb_output = sb.read_notebook(OUTPUT_NOTEBOOK)
|
||||
assert len(nb_output.scraps["aps"].data) == 2
|
||||
assert nb_output.scraps["num_test_images"].data == 38
|
||||
|
||||
|
||||
@pytest.mark.gpu
|
||||
@pytest.mark.notebooks
|
||||
def test_12_notebook_run(
|
||||
|
|
|
|||
|
|
@ -12,6 +12,7 @@ import shutil
|
|||
import urllib.request
|
||||
import xml.etree.ElementTree as ET
|
||||
|
||||
from .references.anno_coco2voc import coco2voc_main
|
||||
|
||||
class Urls:
|
||||
# for now hardcoding base url into Urls class
|
||||
|
|
@ -158,6 +159,24 @@ def coco_labels() -> List[str]:
|
|||
]
|
||||
|
||||
|
||||
def coco2voc(
|
||||
anno_path: str,
|
||||
output_dir: str,
|
||||
anno_type: str = "instance",
|
||||
download_images: bool = False
|
||||
) -> None:
|
||||
""" Convert COCO annotation (single .json file) to Pascal VOC annotations
|
||||
(multiple .xml files).
|
||||
|
||||
Args:
|
||||
anno_path: path to coco-formated .json annotation file
|
||||
output_dir: root output directory
|
||||
anno_type: "instance" for rectangle annotation, or "keypoint" for keypoint annotation.
|
||||
download_images: if true then download images from their urls.
|
||||
"""
|
||||
coco2voc_main(anno_path, output_dir, anno_type, download_images)
|
||||
|
||||
|
||||
def extract_masks_from_labelbox_json(
|
||||
labelbox_json_path: Union[str, Path],
|
||||
data_dir: Union[str, Path],
|
||||
|
|
|
|||
|
|
@ -233,6 +233,7 @@ class DetectionDataset:
|
|||
seed: int = None,
|
||||
allow_negatives: bool = False,
|
||||
labels: List[str] = None,
|
||||
max_num_images = None,
|
||||
):
|
||||
""" initialize dataset
|
||||
|
||||
|
|
@ -270,6 +271,7 @@ class DetectionDataset:
|
|||
self.seed = seed
|
||||
self.keypoint_meta = keypoint_meta
|
||||
self.labels = labels
|
||||
self.max_num_images = max_num_images
|
||||
|
||||
# read annotations
|
||||
self._read_annos()
|
||||
|
|
@ -305,6 +307,7 @@ class DetectionDataset:
|
|||
# it's assumed that the annotation filenames end with .xml.
|
||||
# If im_dir is not provided, then the image paths are read from inside
|
||||
# the .xml annotations.
|
||||
im_paths = None
|
||||
if self.im_dir is None:
|
||||
anno_filenames = sorted(os.listdir(self.root / self.anno_dir))
|
||||
else:
|
||||
|
|
@ -316,6 +319,13 @@ class DetectionDataset:
|
|||
os.path.splitext(s)[0] + ".xml" for s in im_filenames
|
||||
]
|
||||
|
||||
# Reduce number of images if max_num_images is set
|
||||
if self.max_num_images and len(anno_filenames) > self.max_num_images:
|
||||
indices = np.unique(np.floor(np.linspace(0, len(anno_filenames)-1, self.max_num_images)).astype(int))
|
||||
anno_filenames = [anno_filenames[i] for i in indices]
|
||||
if im_paths:
|
||||
im_paths = [im_paths[i] for i in indices]
|
||||
|
||||
# Read all annotations
|
||||
self.im_paths = []
|
||||
self.anno_paths = []
|
||||
|
|
|
|||
|
|
@ -27,6 +27,7 @@ import matplotlib.pyplot as plt
|
|||
|
||||
from .references.engine import train_one_epoch, evaluate
|
||||
from .references.coco_eval import CocoEvaluator
|
||||
from .references.pycocotools_cocoeval import compute_ap
|
||||
from .bbox import bboxes_iou, DetectionBbox
|
||||
from ..common.gpu import torch_device
|
||||
|
||||
|
|
@ -256,36 +257,24 @@ def get_pretrained_keypointrcnn(
|
|||
|
||||
|
||||
def _calculate_ap(
|
||||
e: CocoEvaluator, iou_threshold_idx: Union[int, slice] = slice(0, None)
|
||||
e: CocoEvaluator,
|
||||
iou_thres: float = None,
|
||||
area_range: str ='all',
|
||||
max_detections: int = 100,
|
||||
mode: int = 1,
|
||||
) -> Dict[str, float]:
|
||||
""" Calculate the Average Precision (AP) by averaging all iou
|
||||
thresholds across all labels.
|
||||
""" Calculate the average precision/recall for differnt IoU ranges.
|
||||
|
||||
coco_eval.eval['precision'] is a 5-dimensional array. Each dimension
|
||||
represents the following:
|
||||
1. [T] 10 evenly distributed thresholds for IoU, from 0.5 to 0.95. By
|
||||
default, we use slice(0, None) which is the average from 0.5 to 0.95.
|
||||
2. [R] 101 recall thresholds, from 0 to 101
|
||||
3. [K] label, set to slice(0, None) to get precision over all the labels in
|
||||
the dataset. Then take the mean over all labels.
|
||||
4. [A] area size range of the target (all-0, small-1, medium-2, large-3)
|
||||
5. [M] The maximum number of detection frames in a single image where index
|
||||
0 represents max_det=1, 1 represents max_det=10, 2 represents max_det=100
|
||||
|
||||
Therefore, coco_eval.eval['precision'][0, :, 0, 0, 2] represents the value
|
||||
of 101 precisions corresponding to 101 recalls from 0 to 100 when IoU=0.5.
|
||||
Args:
|
||||
iou_thres: IoU threshold (options: value in [0.5, 0.55, 0.6, ..., 0.95] or None to average over that range)
|
||||
area_range: area size range of the target (options: ['all', 'small', 'medium', 'large'])
|
||||
max_detections: maximum number of detection frames in a single image (options: [1, 10, 100])
|
||||
mode: set to 1 for average precision and otherwise returns average recall
|
||||
"""
|
||||
precision_settings = (
|
||||
iou_threshold_idx,
|
||||
slice(0, None),
|
||||
slice(0, None),
|
||||
0,
|
||||
2,
|
||||
)
|
||||
ap = {
|
||||
k: np.mean(np.mean(v.eval["precision"][precision_settings]))
|
||||
for k, v in e.coco_eval.items()
|
||||
}
|
||||
ap = {}
|
||||
for key in e.coco_eval:
|
||||
ap[key] = compute_ap(e.coco_eval[key], iouThr=iou_thres, areaRng=area_range, maxDets=max_detections, ap=mode)
|
||||
|
||||
return ap
|
||||
|
||||
|
||||
|
|
@ -554,7 +543,7 @@ class DetectionLearner:
|
|||
e = self.evaluate(dl=self.dataset.test_dl)
|
||||
self.ap.append(_calculate_ap(e))
|
||||
self.ap_iou_point_5.append(
|
||||
_calculate_ap(e, iou_threshold_idx=0)
|
||||
_calculate_ap(e)
|
||||
)
|
||||
|
||||
def plot_precision_loss_curves(
|
||||
|
|
|
|||
|
|
@ -0,0 +1,214 @@
|
|||
# Code copied and slightly modified from:
|
||||
# https://github.com/CasiaFan/Dataset_to_VOC_converter/blob/master/anno_coco2voc.py
|
||||
#
|
||||
# Most modifications are hlighlighted by the keyword "EDITED".
|
||||
|
||||
|
||||
import argparse, json
|
||||
import cytoolz
|
||||
from lxml import etree, objectify
|
||||
import os, re
|
||||
import urllib.request, pdb
|
||||
from urllib.parse import urlparse
|
||||
|
||||
|
||||
def instance2xml_base(anno, download_images):
|
||||
# EDITED - make coco_url optional since only used when downloading the images
|
||||
if 'coco_url' not in anno:
|
||||
if 'url' in anno:
|
||||
anno['coco_url'] = anno['url']
|
||||
elif not download_images:
|
||||
anno['coco_url'] = "not used anywhere in code"
|
||||
else:
|
||||
raise Exception("Annotation has to contain a 'url' or 'coco_url' field to download the image.")
|
||||
|
||||
E = objectify.ElementMaker(annotate=False)
|
||||
anno_tree = E.annotation(
|
||||
E.folder('VOC2014_instance/{}'.format(anno['category_id'])),
|
||||
E.filename(anno['file_name']),
|
||||
E.source(
|
||||
E.database('MS COCO 2014'),
|
||||
E.annotation('MS COCO 2014'),
|
||||
E.image('Flickr'),
|
||||
E.url(anno['coco_url'])
|
||||
),
|
||||
E.size(
|
||||
E.width(anno['width']),
|
||||
E.height(anno['height']),
|
||||
E.depth(3)
|
||||
),
|
||||
E.segmented(0),
|
||||
)
|
||||
return anno_tree
|
||||
|
||||
|
||||
def instance2xml_bbox(anno, bbox_type='xyxy'):
|
||||
"""bbox_type: xyxy (xmin, ymin, xmax, ymax); xywh (xmin, ymin, width, height)"""
|
||||
assert bbox_type in ['xyxy', 'xywh']
|
||||
if bbox_type == 'xyxy':
|
||||
xmin, ymin, w, h = anno['bbox']
|
||||
xmax = xmin+w
|
||||
ymax = ymin+h
|
||||
else:
|
||||
xmin, ymin, xmax, ymax = anno['bbox']
|
||||
E = objectify.ElementMaker(annotate=False)
|
||||
anno_tree = E.object(
|
||||
E.name(anno['category_id']),
|
||||
E.bndbox(
|
||||
E.xmin(xmin),
|
||||
E.ymin(ymin),
|
||||
E.xmax(xmax),
|
||||
E.ymax(ymax)
|
||||
),
|
||||
E.difficult(anno['iscrowd'])
|
||||
)
|
||||
return anno_tree
|
||||
|
||||
|
||||
def parse_instance(content, outdir, download_images = False):
|
||||
categories = {d['id']: d['name'] for d in content['categories']}
|
||||
|
||||
# EDITED - make sure image_id is of type int (and not of type string)
|
||||
for i in range(len(content['annotations'])):
|
||||
content['annotations'][i]['image_id'] = int(content['annotations'][i]['image_id'])
|
||||
|
||||
# EDITED - save all annotation .xml files into same sub-directory
|
||||
anno_dir = os.path.join(outdir, "annotations")
|
||||
if not os.path.exists(anno_dir):
|
||||
os.makedirs(anno_dir)
|
||||
|
||||
# EDITED - download images
|
||||
if download_images:
|
||||
im_dir = os.path.join(outdir, "images")
|
||||
if not os.path.exists(im_dir):
|
||||
os.makedirs(im_dir)
|
||||
|
||||
for index, obj in enumerate(content['images']):
|
||||
print(f"Downloading image {index} of {len(content['images'])} from: {obj['coco_url']}")
|
||||
|
||||
# Update 'filename' field to be a (local) filename and not a url
|
||||
im_local_filename = os.path.splitext(os.path.basename(obj['file_name']))[0] + ".jpg"
|
||||
obj['file_name'] = im_local_filename
|
||||
|
||||
# download image
|
||||
dst_path = os.path.join(im_dir, im_local_filename)
|
||||
urllib.request.urlretrieve(obj['coco_url'], dst_path)
|
||||
|
||||
# merge images and annotations: id in images vs image_id in annotations
|
||||
merged_info_list = list(map(cytoolz.merge, cytoolz.join('id', content['images'], 'image_id', content['annotations'])))
|
||||
|
||||
# convert category id to name
|
||||
for instance in merged_info_list:
|
||||
assert 'category_id' in instance, f"WARNING: annotation error: image {instance['file_name']} has a rectangle without a 'category_id' field."
|
||||
instance['category_id'] = categories[instance['category_id']]
|
||||
|
||||
# group by filename to pool all bbox in same file
|
||||
img_filenames = {}
|
||||
names_groups = cytoolz.groupby('file_name', merged_info_list).items()
|
||||
for index, (name, groups) in enumerate(names_groups):
|
||||
print(f"Converting annotations for image {index} of {len(names_groups)}: {name}")
|
||||
assert not name.lower().startswith(("http:","https:")), "Image seems to be a url rather than local. Need to set 'download_images' = False"
|
||||
|
||||
anno_tree = instance2xml_base(groups[0], download_images)
|
||||
# if one file have multiple different objects, save it in each category sub-directory
|
||||
filenames = []
|
||||
for group in groups:
|
||||
filename = os.path.splitext(name)[0] + ".xml"
|
||||
|
||||
# EDITED - save all annotations in single folder, rather than separate folders for each object
|
||||
#filenames.append(os.path.join(outdir, re.sub(" ", "_", group['category_id']), filename))
|
||||
filenames.append(os.path.join(anno_dir, filename))
|
||||
|
||||
anno_tree.append(instance2xml_bbox(group, bbox_type='xyxy'))
|
||||
|
||||
for filename in filenames:
|
||||
etree.ElementTree(anno_tree).write(filename, pretty_print=True)
|
||||
|
||||
|
||||
def keypoints2xml_base(anno):
|
||||
annotation = etree.Element("annotation")
|
||||
etree.SubElement(annotation, "folder").text = "VOC2014_keypoints"
|
||||
etree.SubElement(annotation, "filename").text = anno['file_name']
|
||||
source = etree.SubElement(annotation, "source")
|
||||
etree.SubElement(source, "database").text = "MS COCO 2014"
|
||||
etree.SubElement(source, "annotation").text = "MS COCO 2014"
|
||||
etree.SubElement(source, "image").text = "Flickr"
|
||||
etree.SubElement(source, "url").text = anno['coco_url']
|
||||
size = etree.SubElement(annotation, "size")
|
||||
etree.SubElement(size, "width").text = str(anno["width"])
|
||||
etree.SubElement(size, "height").text = str(anno["height"])
|
||||
etree.SubElement(size, "depth").text = '3'
|
||||
etree.SubElement(annotation, "segmented").text = '0'
|
||||
return annotation
|
||||
|
||||
|
||||
def keypoints2xml_object(anno, xmltree, keypoints_dict, bbox_type='xyxy'):
|
||||
assert bbox_type in ['xyxy', 'xywh']
|
||||
if bbox_type == 'xyxy':
|
||||
xmin, ymin, w, h = anno['bbox']
|
||||
xmax = xmin+w
|
||||
ymax = ymin+h
|
||||
else:
|
||||
xmin, ymin, xmax, ymax = anno['bbox']
|
||||
key_object = etree.SubElement(xmltree, "object")
|
||||
etree.SubElement(key_object, "name").text = anno['category_id']
|
||||
bndbox = etree.SubElement(key_object, "bndbox")
|
||||
etree.SubElement(bndbox, "xmin").text = str(xmin)
|
||||
etree.SubElement(bndbox, "ymin").text = str(ymin)
|
||||
etree.SubElement(bndbox, "xmax").text = str(xmax)
|
||||
etree.SubElement(bndbox, "ymax").text = str(ymax)
|
||||
etree.SubElement(key_object, "difficult").text = '0'
|
||||
keypoints = etree.SubElement(key_object, "keypoints")
|
||||
for i in range(0, len(keypoints_dict)):
|
||||
keypoint = etree.SubElement(keypoints, keypoints_dict[i+1])
|
||||
etree.SubElement(keypoint, "x").text = str(anno['keypoints'][i*3])
|
||||
etree.SubElement(keypoint, "y").text = str(anno['keypoints'][i*3+1])
|
||||
etree.SubElement(keypoint, "v").text = str(anno['keypoints'][i*3+2])
|
||||
return xmltree
|
||||
|
||||
|
||||
def parse_keypoints(content, outdir):
|
||||
keypoints = dict(zip(range(1, len(content['categories'][0]['keypoints'])+1), content['categories'][0]['keypoints']))
|
||||
# merge images and annotations: id in images vs image_id in annotations
|
||||
merged_info_list = map(cytoolz.merge, cytoolz.join('id', content['images'], 'image_id', content['annotations']))
|
||||
# convert category name to person
|
||||
for keypoint in merged_info_list:
|
||||
keypoint['category_id'] = "person"
|
||||
# group by filename to pool all bbox and keypoint in same file
|
||||
for name, groups in cytoolz.groupby('file_name', merged_info_list).items():
|
||||
filename = os.path.join(outdir, os.path.splitext(name)[0]+".xml")
|
||||
anno_tree = keypoints2xml_base(groups[0])
|
||||
for group in groups:
|
||||
anno_tree = keypoints2xml_object(group, anno_tree, keypoints, bbox_type="xyxy")
|
||||
doc = etree.ElementTree(anno_tree)
|
||||
doc.write(open(filename, "w"), pretty_print=True)
|
||||
print("Formating keypoints xml file {} done!".format(name))
|
||||
|
||||
|
||||
def coco2voc_main(anno_file, output_dir, anno_type, download_images = False):
|
||||
if not os.path.exists(output_dir):
|
||||
os.makedirs(output_dir)
|
||||
content = json.load(open(anno_file, 'r'))
|
||||
|
||||
if anno_type == 'instance':
|
||||
# EDITED - save all annotations in single folder, rather than separate folders for each object
|
||||
# make subdirectories
|
||||
# sub_dirs = [re.sub(" ", "_", cate['name']) for cate in content['categories']] #EDITED
|
||||
# for sub_dir in sub_dirs:
|
||||
# sub_dir = os.path.join(output_dir, str(sub_dir))
|
||||
# if not os.path.exists(sub_dir):
|
||||
# os.makedirs(sub_dir)
|
||||
parse_instance(content, output_dir, download_images)
|
||||
elif anno_type == 'keypoint':
|
||||
parse_keypoints(content, output_dir)
|
||||
else:
|
||||
error
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
parser = argparse.ArgumentParser()
|
||||
parser.add_argument("--anno_file", help="annotation file for object instance/keypoint")
|
||||
parser.add_argument("--type", type=str, help="object instance or keypoint", choices=['instance', 'keypoint'])
|
||||
parser.add_argument("--output_dir", help="output directory for voc annotation xml file")
|
||||
args = parser.parse_args()
|
||||
main(args.anno_file, args.output_dir, args.type)
|
||||
|
|
@ -0,0 +1,41 @@
|
|||
# Code copied and slighly re-arranged from:
|
||||
# https://github.com/cocodataset/cocoapi/blob/master/PythonAPI/pycocotools/cocoeval.py#L422
|
||||
|
||||
import numpy as np
|
||||
from pycocotools.cocoeval import Params
|
||||
|
||||
|
||||
def compute_ap(v, ap=1, iouThr=None, areaRng='all', maxDets=100, verbose = False):
|
||||
p = Params()
|
||||
aind = [i for i, aRng in enumerate(p.areaRngLbl) if aRng == areaRng]
|
||||
mind = [i for i, mDet in enumerate(p.maxDets) if mDet == maxDets]
|
||||
if ap == 1:
|
||||
# dimension of precision: [TxRxKxAxM]
|
||||
s = v.eval['precision']
|
||||
# IoU
|
||||
if iouThr is not None:
|
||||
t = np.where(iouThr == p.iouThrs)[0]
|
||||
s = s[t]
|
||||
s = s[:,:,:,aind,mind]
|
||||
else:
|
||||
# dimension of recall: [TxKxAxM]
|
||||
s = self.eval['recall']
|
||||
if iouThr is not None:
|
||||
t = np.where(iouThr == p.iouThrs)[0]
|
||||
s = s[t]
|
||||
s = s[:,:,aind,mind]
|
||||
if len(s[s>-1])==0:
|
||||
mean_s = -1
|
||||
else:
|
||||
mean_s = np.mean(s[s>-1])
|
||||
|
||||
if verbose:
|
||||
iStr = ' {:<18} {} @[ IoU={:<9} | area={:>6s} | maxDets={:>3d} ] = {:0.3f}'
|
||||
titleStr = 'Average Precision' if ap == 1 else 'Average Recall'
|
||||
typeStr = '(AP)' if ap==1 else '(AR)'
|
||||
iouStr = '{:0.2f}:{:0.2f}'.format(p.iouThrs[0], p.iouThrs[-1]) \
|
||||
if iouThr is None else '{:0.2f}'.format(iouThr)
|
||||
resStr = iStr.format(titleStr, typeStr, iouStr, areaRng, maxDets, mean_s)
|
||||
print(resStr)
|
||||
|
||||
return mean_s
|
||||
Загрузка…
Ссылка в новой задаче