Input tiling#
This tutorial will show you how to tile the input to a model, using the Tiler
.
Warning
This tutorial assumes that you have already installed anomalib. If not, please refer to the Installation section.
Warning
Only selected models support tiling. In the current version of Anomalib, these are:
General tiling information#
The general idea of input tiling is that the image is split into a rectangular grid of tiles as a pre-processing step, usually in order to reduce memory usage. By passing individual tiles to the model as input instead of full images, tiling reduces the model’s input dimensions, while maintaining the effective input resolution of the images content-wise.
Note
Tiler in Anomalib by default stacks the tiles batch-wise, so the memory consumption stays unchanged if the batch size is not reduced.
The process of tiling is parametrized by four parameters tile_size
, stride
, remove_border_count
, and mode
.
tile_size
- determines the size of our tiles. Can be either a single number (square tiles) or a tuple.stride
- determines by how much we move in each direction when “cutting” the image into tiles. Can be either a single number (same step in both directions) or a tuple.remove_border_count
- how many pixels are removed at the border of the image before tiling (defaults to 0).mode
- what type of upscaling is used when the image isn’t exactly divisible into tile-set specified by the parameterstile_size
andstride
(defaults to padding).
In most cases, we are only interested in the first two parameters - tile_size
and stride
. For the other two, refer to Tiler implementation.
Tiling setup#
We can utilize the tiling in two ways. Either with the CLI or by using the API.
In both cases, we need to use the TilerConfigurationCallback
.
This callback is responsible for assigning the tiler object to the model before the training starts.
Note
Besides the arguments from Tiler
, TilerConfigurationCallback
also has an additional enable
argument, which must be set to True
if we want the tiling to happen.
To use tiling from the API, we need to initialize the TilerConfigurationCallback
and pass it to the engine:
1# Import the required modules
2from anomalib.data import MVTec
3from anomalib.engine import Engine
4from anomalib.models import Padim
5from anomalib.callbacks import TilerConfigurationCallback
6
7# Initialize the datamodule and model
8datamodule = MVTec(num_workers=0, image_size=(128, 128))
9model = Padim()
10
11# prepare tiling configuration callback
12tiler_config_callback = TilerConfigurationCallback(enable=True, tile_size=[128, 64], stride=64)
13
14# pass the tiling configuration callback to the engine
15engine = Engine(image_metrics=["AUROC"], pixel_metrics=["AUROC"], callbacks=[tiler_config_callback])
16
17# train the model (tiling is seamlessly utilized in the background)
18engine.fit(datamodule=datamodule, model=model)
Using CLI arguments
We can set the TilerConfigurationCallback
and its init arguments directly from the CLI.
We pass it as trainer.callback, and then provide the parameters:
anomalib train --model Padim --data anomalib.data.MVTec
--trainer.callbacks anomalib.callbacks.tiler_configuration.TilerConfigurationCallback
--trainer.callbacks.enable True
--trainer.callbacks.tile_size 128
--trainer.callbacks.stride 64
Using config
For more advanced configuration, we can prepare the config file:
1trainer.callbacks:
2 class_path: anomalib.callbacks.tiler_configuration.TilerConfigurationCallback
3 init_args:
4 enable: True
5 tile_size: [128, 256]
6 stride: 64
Then use the config from the CLI:
anomalib train --model Padim --data anomalib.data.MVTec --config config.yaml