Object detection: TypeError: __call__() takes 2 positional arguments but 3 were given

vision
1

Hi i am creating an object detector, for that I am using the following code for creating the custom dataset:

class MIMICCXRDataset(Dataset):

def __init__(self,
             split,
             args,
             radgraph_sids,
             radgraph_adj_mtx,
             mode,
             transform=None):

    self.args = args
    self.mode = mode
    self.transform = transform
 ..........

def __getitem__(self, idx):
    # 1. get the sample key, i.e., DICOM ID
    dicom_id = self.dicom_ids[idx]
    study_id = list(self.df_master_sel.loc[self.df_master_sel['dicom_id'] == dicom_id, 'study_id'])[0]
    idx_sid = list(self.arr_radgraph_sids).index(str(study_id))

    # 2. select and load image
    img_pth_old = self.df_master_sel.iloc[idx, 4]
    img_pth_components = img_pth_old.split("/")
    img_pth = os.path.join(
        self.args.image_path_ocean_shared,
        *img_pth_components[11:]
    )

    image = Image.open(img_pth).convert('RGB')
    np_res_bbox = np.array(res_bbox)

    if len(np_res_bbox.shape) == 1:
        np_res_bbox = np.reshape(np_res_bbox, (1, -1))

    for i in range(np_res_bbox.shape[0]):
        x1 = np_res_bbox[i, 0]
        y1 = np_res_bbox[i, 1]
        x2 = np_res_bbox[i, 2]
        y2 = np_res_bbox[i, 3]
        boxes.append([x1, y1, x2, y2])

    boxes = torch.as_tensor(boxes, dtype=torch.float32)
    if disease == "pneumonia":
        labels = torch.ones((boxes.size(0),), dtype=torch.int64)
    elif disease == "pneumothorax":
        labels = torch.ones((boxes.size(0),), dtype=torch.int64)

    area = (boxes[:, 3] - boxes[:, 1]) * (boxes[:, 2] - boxes[:, 0])
    iscrowd = torch.zeros((boxes.size(0),), dtype=torch.int64)
    target = {}
    target["boxes"] = boxes
    target["labels"] = labels
    target["image_id"] = torch.tensor(idx)
    target["area"] = area
    target["iscrowd"] = iscrowd

    if self.transform is not None:
        image, target = self.transform(image, target)

I am getting the error in the transforms. My transforms are as follows:
train_dataset = MIMICCXRDataset(
split=split,
args=args,
radgraph_sids=arr_radgraph_sids,
radgraph_adj_mtx=arr_radgraph_adj,
mode=‘bbox’,
transform=transforms.Compose([
transforms.Resize(args.resize),
# resize smaller edge to args.resize and the
# aspect ratio the same for the longer edge
transforms.CenterCrop(args.resize),
transforms.ToTensor(), # convert pixel value to [0, 1]
normalize
])
)

2

The used transformations accept a single input only while you are passing the image and target to them.
To allow two inputs, you could refer to the Object Detection tutorial and use the util. functions from references/detection/transforms.py.