When inference, do I need to convert image to jpg format first?

[wennyHou]

When inference, do I need to convert image to jpg format first? Thanks

[CauchyComplete]

CAT-Net requires a JPEG image as input but the inference code automatically converts a non-JPEG image to JPEG at the very beginning.

[FathUMinUllah3797]

@CauchyComplete Do we need to convert all the images into jpg before training? If yes, why there are still images with .tiff and .png extension in the lists you provided. Please response to this query as I am getting the following error.

[LIBJPEG ERROR]: Not a JPEG file: starts with 0x49 0x49
[LIBJPEG] Error occurs during reading file.
Traceback (most recent call last):
File "D:\ImageManipulation\CAT-Net-main\CAT-Net-main\tools\train.py", line 233, in
main()
File "D:\ImageManipulation\CAT-Net-main\CAT-Net-main\tools\train.py", line 185, in main
trainloader, optimizer, model, writer_dict, final_output_dir)
File "D:\ImageManipulation\CAT-Net-main\CAT-Net-main\tools..\lib\core\function.py", line 58, in train
for i_iter, (images, labels, qtable) in enumerate(trainloader):
File "C:\Users\MSI\anaconda3\envs\cat\lib\site-packages\torch\utils\data\dataloader.py", line 521, in next
data = self._next_data()
File "C:\Users\MSI\anaconda3\envs\cat\lib\site-packages\torch\utils\data\dataloader.py", line 561, in _next_data
data = self._dataset_fetcher.fetch(index) # may raise StopIteration
File "C:\Users\MSI\anaconda3\envs\cat\lib\site-packages\torch\utils\data_utils\fetch.py", line 49, in fetch
data = [self.dataset[idx] for idx in possibly_batched_index]
File "C:\Users\MSI\anaconda3\envs\cat\lib\site-packages\torch\utils\data_utils\fetch.py", line 49, in
data = [self.dataset[idx] for idx in possibly_batched_index]
File "D:\ImageManipulation\CAT-Net-main\CAT-Net-main\tools..\Splicing\data\data_core.py", line 99, in getitem
return self.dataset_list[index//self.smallest].get_tamp(index%self.smallest)
File "D:\ImageManipulation\CAT-Net-main\CAT-Net-main\tools..\Splicing\data\dataset_CASIA.py", line 55, in get_tamp
return self._create_tensor(tamp_path, mask)
File "D:\ImageManipulation\CAT-Net-main\CAT-Net-main\tools..\Splicing\data\AbstractDataset.py", line 111, in _create_tensor
DCT_coef, qtables = self._get_jpeg_info(im_path)
File "D:\ImageManipulation\CAT-Net-main\CAT-Net-main\tools..\Splicing\data\AbstractDataset.py", line 44, in _get_jpeg_info
jpeg = jpegio.read(str(im_path))
File "C:\Users\MSI\anaconda3\envs\cat\lib\site-packages\jpegio\io.py", line 10, in read
obj.read(fpath)
File "jpegio\decompressedjpeg.pyx", line 62, in jpegio.decompressedjpeg.DecompressedJpeg.read
File "jpegio\decompressedjpeg.pyx", line 80, in jpegio.decompressedjpeg.DecompressedJpeg.read
OSError: An error has occurs in reading the file.
[LIBJPEG ERROR]: Improper call to JPEG library in state 0

[CauchyComplete]

In the inference dataset(dataset='arbitrary'), non-JPEG images are automatically converted to JPEG files and plugged into the model:

CAT-Net/tools/infer.py

Line 75 in f1716b0

test_dataset = splicing_dataset(crop_size=None, grid_crop=True, blocks=('RGB', 'DCTvol', 'qtable'), DCT_channels=1, mode='arbitrary', read_from_jpeg=True) # full model

During training, we should manually convert them into JPEG files.

I implemented it like this for speed-up training because training uses converted files multiple times.