Confused of your inference mode about flownet2-pytorch

[fitsumreda]

Hi @qijiezhao , ImagesFromFolder is the dataloader for regular images for simplistic test purpose, it expects two input arguments
root - folder path that contains sequentially named RGB images
iext - an extension for the image files
both of these can be passed in as arguments, inference_dataset_root and inference_dataset_iext.
Could you explain a little about your custom dataset?
Any way ImagesFromFolder can be extended?

[qijiezhao]

Hi @fitsumreda , Thx for your reply, here I modified some code in main.py and it is approaching worked out:
```

if exists(args.inference_dataset_root):
            #inference_dataset = args.inference_dataset_class(args, False, **tools.kwargs_from_args(args, 'inference_dataset'))
            #block.log('Inference Dataset: {}'.format(args.inference_dataset))
            #block.log('Inference Input: {}'.format(' '.join([str([d for d in x.size()]) for x in inference_dataset[0][0]])))
            #block.log('Inference Targets: {}'.format(' '.join([str([d for d in x.size()]) for x in inference_dataset[0][1]])))
            #inference_loader = DataLoader(inference_dataset, batch_size=args.inference_batch_size, shuffle=False, **gpuargs)
            inference_loader = DataLoader(
            
datasets.ImagesFromFolder(args=args,is_cropped=False,root=args.inference_dataset_root,iext='jpg'),
                batch_size=args.inference_batch_size,shuffle=False,**gpuargs
            )

detailed __getitem__ function could be implemented ,as you said...

Now, without checkpoints file,  everything is nearly OK xi:-D

[sklf]

@fitsumreda @qijiezhao Hello, I read your discussion. And I used the code as above. It is as follows:

if exists(args.inference_dataset_root):
        inference_loader = DataLoader(datasets.ImagesFromFolder(args=args, is_cropped=False, root=args.inference_dataset_root, iext='png'),
            batch_size =args.effective_inference_batch_size, shuffle=False, **inf_gpuargs)

and my custom dataset has 100 '*.png' images. However, I ran the model Flownet2S with the following code

python main.py --inference --model FlowNet2S --save_flow --inference_dataset_root ~/files/Dataset/posetrack/Flowimgs/imgs/val/000342_mpii_test/ --resume ./work/FlowNet2-S_checkpoint.pth.tar 

and found I only got 24 flo file. I used 3 GPUs to ran it. So can you tell me why I can't get all flo files? Thanks!