Add Normalization to DataBlock

Part 1 (2020)
1

Hello,
I am working on Object Detection task.
How can I add Normalization from stats to DataBlock batch_tfms?

If I use:

mean = [0.485, 0.456, 0.406]
std =  [0.229, 0.224, 0.225]
coco = DataBlock(blocks=(ImageBlock, BBoxBlock, BBoxLblBlock),
                 get_items=get_image_files,
                 splitter=RandomSplitter(valid_pct=0.0),
                 get_y=[lambda o: img2bbox[o.name][0], lambda o: img2bbox[o.name][1]], 
                 item_tfms=Resize(TEMP_figsize),
                 batch_tfms=[IntToFloatTensor(), Normalize.from_stats(mean,std), aug_transforms()],
                 n_inp=1)

I will get a warning:
Could not do one pass in your dataloader, there is something wrong in it

And error:
‘list’ object is not callable

How to add normalization properly, as I do not see it in default batch_tfms?
Thank you!

2

The solution would be:

coco = DataBlock(blocks=(ImageBlock, BBoxBlock, BBoxLblBlock),
                 get_items=get_image_files,
                 splitter=RandomSplitter(valid_pct=0.0),
                 get_y=[lambda o: img2bbox[o.name][0], lambda o: img2bbox[o.name][1]], 
                 item_tfms=Resize(TEMP_figsize),
                 batch_tfms=[Normalize.from_stats(*imagenet_stats), *aug_transforms()],
                 n_inp=1)
2 Likes