WZMIAOMIAO/deep-learning-for-image-processing · error · ValueError

Image '{}' format not JPEG

Error message

Image '{}' format not JPEG

What it means

VOCDataSet.__getitem__ opens the image named in the annotation XML and requires its format to be exactly 'JPEG' (PIL's format attribute). Any other format — PNG, BMP, GIF, MPO, or a JPEG whose PIL format reads differently — raises ValueError with the image path. The code assumes a standard VOC2012 dataset where all images are JPEG.

Source

Thrown at pytorch_object_detection/faster_rcnn/my_dataset.py:72

        with open(json_file, 'r') as f:
            self.class_dict = json.load(f)

        self.transforms = transforms

    def __len__(self):
        return len(self.xml_list)

    def __getitem__(self, idx):
        # read xml
        xml_path = self.xml_list[idx]
        with open(xml_path) as fid:
            xml_str = fid.read()
        xml = etree.fromstring(xml_str)
        data = self.parse_xml_to_dict(xml)["annotation"]
        img_path = os.path.join(self.img_root, data["filename"])
        image = Image.open(img_path)
        if image.format != "JPEG":
            raise ValueError("Image '{}' format not JPEG".format(img_path))

        boxes = []
        labels = []
        iscrowd = []
        assert "object" in data, "{} lack of object information.".format(xml_path)
        for obj in data["object"]:
            xmin = float(obj["bndbox"]["xmin"])
            xmax = float(obj["bndbox"]["xmax"])
            ymin = float(obj["bndbox"]["ymin"])
            ymax = float(obj["bndbox"]["ymax"])

            # 进一步检查数据，有的标注信息中可能有w或h为0的情况，这样的数据会导致计算回归loss为nan
            if xmax <= xmin or ymax <= ymin:
                print("Warning: in '{}' xml, there are some bbox w/h <=0".format(xml_path))
                continue
            
            boxes.append([xmin, ymin, xmax, ymax])
            labels.append(self.class_dict[obj["name"]])

View on GitHub (pinned to 1ec3fe6f37)

Solutions

  1. Convert all dataset images to JPEG: e.g. `for f in *.png; do convert $f ${f%.png}.jpg; done` and update the XML filenames.
  2. Check data['filename'] in the failing XML matches an actual .jpg file with JPEG content.
  3. Re-encode with PIL: Image.open(p).convert('RGB').save(p, 'JPEG').
  4. Remove or fix non-conforming samples from the XML/ImageSets lists.
  5. Relax the check (image.convert('RGB')) if you intentionally support other formats.

Example fix

# before
image = Image.open(img_path)
if image.format != "JPEG":
    raise ValueError(...)
# after
image = Image.open(img_path)
if image.format != "JPEG":
    image = image.convert("RGB")  # tolerate PNG/BMP inputs
Defensive patterns

Strategy: validation

Validate before calling

from PIL import Image
img_path = os.path.join(self.img_root, data['filename'])
with Image.open(img_path) as im:
    assert im.format == 'JPEG', f"{img_path} is {im.format}, convert to JPEG"

Type guard

def is_jpeg(path: str) -> bool:
    try:
        with Image.open(path) as im:
            return im.format == 'JPEG'
    except Exception:
        return False

Try / catch

try:
    image, target = dataset[i]
except ValueError as e:
    print(f'Skipping non-JPEG sample: {e}')
    continue

Prevention

When it happens

Trigger: The XML annotation's <filename> points to a non-JPEG file (e.g. image.png/image.bmp), a file with a wrong/misleading name, or a corrupted image PIL cannot decode as JPEG.

Common situations: Mixing a custom dataset into VOC layout, images converted/re-encoded with wrong extensions, downloading images that are PNG but renamed .jpg, or camera images with MPO format.

Related errors

AI-assisted analysis of WZMIAOMIAO/deep-learning-for-image-processing@1ec3fe6f37 (2026-08-30). Data as JSON: /api/errors/80b4273993a477b6. Report an issue: GitHub.