Can anyone help me with an example on how to create databunch for object detection with fast.ai-V1 ? I am aware that this will be explained in part -2, but I am asking it out of enthusiaism . Please let me know if anyone has any idea.
The databunch is really just a wrapper around train/valid/test dataloaders. The bits that are different for image segmentation are the datasets which have to include not just image/category but also bbox and possibly mask. A dataset is just an abstract class so you can define getitem to return anything you want.
I have implementation of maskrcnn which uses fastai. Probably could be better e.g. did not use the data block api. And the architecture of maskrcnn is crazy complex. Nevertheless may help.
3 Likes
I want more detailed example using datablock api.
You can find all kind of examples in docs.fast.ai https://docs.fast.ai/data_block.html .There include the Object Detection like below:
data = (ObjectItemList.from_folder(coco)
#Where are the images? -> in coco and its subfolders
.split_by_rand_pct()
#How to split in train/valid? -> randomly with the default 20% in valid
.label_from_func(get_y_func)
#How to find the labels? -> use get_y_func on the file name of the data
.transform(get_transforms(), tfm_y=True)
#Data augmentation? -> Standard transforms; also transform the label images
.databunch(bs=16, collate_fn=bb_pad_collate))
#Finally we convert to a DataBunch, use a batch size of 16,
# and we use bb_pad_collate to collate the data into a mini-batchthis is fine but i want something like how to make a databunch of images and their bounding box values .
See my post here: Object detection databunch