I'm struggling to modify my JSON file. I need to take the following JSON
{"categories": [{"supercategory": "", "id": 0, "name": "Negative", "tiles": 2489, "bboxes": 5527, "className": "Negative"}, {"supercategory": "", "id": 1, "name": "Positive, "tiles": 5227, "bboxes": 15362, "className": "Positive}], "images": [{"id": 224, "file_name": "img1.jpg", "height": 512, "width": 512}, {"id": 225, "file_name": "img2.jpg", "height": 512, "width": 512}], "annotations": [{"id": 716, "image_id": 224, "category_id": 0, "iscrowd": 0, "area": 2856.0, "bbox": [298, 18, 56, 51]}, {"id": 715, "image_id": 224, "category_id": 0, "iscrowd": 0, "area": 4096.0, "bbox": [185, 68, 64, 64]}, {"id": 714, "image_id": 224, "category_id": 0, "iscrowd": 0, "area": 2744.0, "bbox": [354, 10, 56, 49]}, {"id": 717, "image_id": 225, "category_id": 0, "iscrowd": 0, "area": 4096.0, "bbox": [374, 397, 64, 64]}]}
and make this look as following:
[{"image_id":224, "file_name": "img1.jpg", "height": 512, "width": 512, "annotations":[{"bbox":[298.0, 18.0, 56.0, 51.0],"bbox_mode":1,"category_id":"0" }, {"bbox":[185.0, 68.0, 64.0, 64.0],"bbox_mode":1,"category_id":"0" }, {"bbox":[354.0, 10.0, 56.0, 49.0],"bbox_mode":1,"category_id":"0"}]}, {"image_id":225,"file_name":"img2.jpg","height":512,"width":512,"annotations":[{"bbox":[374.0, 397.0, 64.0, 64.0],"bbox_mode":1,"category_id":"0"}]}]
I read the JSON file and loop over the "images" elements, however, I'm stuck in assigning "annotations" for the specific images. As my annotation is in another element my code assigns all the annotation to the first image. I need help to fix this. Here is my code:
json_file = "C:/Temp/python/current_json.json"
elements, element, annotation ={}, {}, {}
listElements, listAnnotations, kk = [], [], []
list = ()
with open(json_file) as f:
dataset_dicts = json.load(f)
for i in dataset_dicts["images"]:
an = []
for a in dataset_dicts['annotations']:
element.update(image_id=int(i["id"]), file_name=i["file_name"], height=int(i["height"]), width=int(i["width"]))
listElements.append(element.copy())
annotation.update(bbox=a['bbox'], bbox_mode=1, category_id=a['category_id'])
listAnnotations.append(an.copy())
element["annotations"] = listAnnotations
kk.append(element)
elements = [element]
with open("python/new_json.json", "w") as f:
f.write(json.dumps(elements))
this code produce the following output:
[{"image_id": 225, "file_name": "img2.jpg", "height": 512, "width": 512, "annotations": [[{"bbox": [298, 18, 56, 51], "bbox_mode": 1, "category_id": 0}, {"bbox": [185, 68, 64, 64], "bbox_mode": 1, "category_id": 0}, {"bbox": [354, 10, 56, 49], "bbox_mode": 1, "category_id": 0}, {"bbox": [374, 397, 64, 64], "bbox_mode": 1, "category_id": 0}], [{"bbox": [298, 18, 56, 51], "bbox_mode": 1, "category_id": 0}, {"bbox": [185, 68, 64, 64], "bbox_mode": 1, "category_id": 0}, {"bbox": [354, 10, 56, 49], "bbox_mode": 1, "category_id": 0}, {"bbox": [374, 397, 64, 64], "bbox_mode": 1, "category_id": 0}]]}]
CodePudding user response：
The problem seems just that you need to filter annotations based on image_id
with open(json_file) as f:
dataset_dicts = json.load(f)
for image in dataset_dicts["images"]:
image["annotations"] = [dict(**a, bbox_mode=1) for a in dataset_dicts['annotations'] if a['image_id'] == image['id']]
with open("python/new_json.json", "w") as f:
json.dump(dataset_dicts["images"], f)
CodePudding user response：
Some errors in your code:
- You don't need to re-iterate over the
annotationsagain and again for every item inimages, remove it out of the loop - You don't need to update all the keys of
elementfor every item inannotations. Remove it out of the loop. - The usage of all other auxiliary containers such as
listElements,listAnnotations, andkkare not needed. You can update the targetelementsdirectly if you refactored your code.
With the points above, consider this approach.
- First, get all the annotations grouped by image_id
- Then, iterate each image. For each image, just get the target annotation from the previous step. Construct the target dictionary and append to the result.
from collections import defaultdict
elements = []
annotation = defaultdict(list)
dataset_dicts = {"categories": [{"supercategory": "", "id": 0, "name": "Negative", "tiles": 2489, "bboxes": 5527, "className": "Negative"}, {"supercategory": "", "id": 1, "name": "Positive", "tiles": 5227, "bboxes": 15362, "className": "Positive"}], "images": [{"id": 224, "file_name": "img1.jpg", "height": 512, "width": 512}, {"id": 225, "file_name": "img2.jpg", "height": 512, "width": 512}], "annotations": [{"id": 716, "image_id": 224, "category_id": 0, "iscrowd": 0, "area": 2856.0, "bbox": [298, 18, 56, 51]}, {"id": 715, "image_id": 224, "category_id": 0, "iscrowd": 0, "area": 4096.0, "bbox": [185, 68, 64, 64]}, {"id": 714, "image_id": 224, "category_id": 0, "iscrowd": 0, "area": 2744.0, "bbox": [354, 10, 56, 49]}, {"id": 717, "image_id": 225, "category_id": 0, "iscrowd": 0, "area": 4096.0, "bbox": [374, 397, 64, 64]}]}
for item in dataset_dicts['annotations']:
annotation[item["image_id"]].append(
{
"bbox": list(map(float, item["bbox"])), # Or just item["bbox"].copy() if you don't intend it to be float
"bbox_mode": 1,
"category_id": str(item["category_id"]), # Or just item["category_id"] if you don't intend it to be string
}
)
for item in dataset_dicts["images"]:
elements.append(
{
"image_id": item["id"],
"file_name": item["file_name"],
"height": item["height"],
"width": item["width"],
"annotations": annotation[item["id"]],
}
)
print(elements)
Output
[
{
"image_id": 224,
"file_name": "img1.jpg",
"height": 512,
"width": 512,
"annotations": [
{
"bbox": [
298.0,
18.0,
56.0,
51.0
],
"bbox_mode": 1,
"category_id": "0"
},
{
"bbox": [
185.0,
68.0,
64.0,
64.0
],
"bbox_mode": 1,
"category_id": "0"
},
{
"bbox": [
354.0,
10.0,
56.0,
49.0
],
"bbox_mode": 1,
"category_id": "0"
}
]
},
{
"image_id": 225,
"file_name": "img2.jpg",
"height": 512,
"width": 512,
"annotations": [
{
"bbox": [
374.0,
397.0,
64.0,
64.0
],
"bbox_mode": 1,
"category_id": "0"
}
]
}
]