UpdatedAugust 12, 2025

Col2lm

Description

The operator rearranges column blocks back into a multidimensional image

Col2Im behaves similarly to PyTorch’s fold https://pytorch.org/docs/stable/generated/torch.nn.Fold.html, but it only supports batched multi-dimensional image tensors. Another implementation in Python with N-dimension support can be found at https://github.com/f-dangel/unfoldNd/.

NOTE: Although specifying image_shape looks redundant because it could be calculated from convolution formulas, it is required as input for more advanced scenarios as explained at PyTorch’s implementation (https://github.com/pytorch/pytorch/blob/master/aten/src/ATen/native/Col2Im.cpp#L10)

Input parameters

specified_outputs_name : array, this parameter lets you manually assign custom names to the output tensors of a node.

Graphs in : cluster, ONNX model architecture.

input (heterogeneous) – T : object, input data tensor to be rearranged from column blocks back into an image. This is a 3-dimensional tensor containing [N, C * n-ary-product(block_shape), L], where N is batch dimension, C is image channel dimension and L is number of blocks.The blocks are enumerated in increasing lexicographic-order of their indices.For example, with an image-size 1020 and block-size 918, there would be 2*3 blocks, enumerated in the order block(0, 0), block(0, 1), block(0, 2), block(1, 0), block(1, 1), block(1, 2).
image_shape (heterogeneous) – tensor(int64) : object, the shape of the spatial dimensions of the image after rearranging the column blocks.This is a 1-dimensional tensor with size of at least 2, containing the value [H_img, W_img] for a 2-D image or [dim_i1, dim_i2, …, dim_iN] for a N-D image.
block_shape (heterogeneous) – tensor(int64) : object, the shape of the block to apply on the input.This is a 1-dimensional tensor of size of at least 2, containing the value [H_block, W_block] for a 2-D image or [dim_b1, dim_b2, …, dim_bN] for a N-D block.This is the block-shape before dilation is applied to it.

Parameters : cluster,

dilations : array, 1-dimensional tensor with dilation value along each spatial axis of the image. If not present, the dilation defaults to 1 along each spatial axis of the image.
Default value “empty”.
pads : array, 1-dimensional tensor with padding value for the beginning and ending along each spatial axis, it can take any value greater than or equal to 0. The value represent the number of pixels added to the beginning and end part of the corresponding axis. pads format should be as follow [x1_begin, x2_begin…x1_end, x2_end,…], where xi_begin is the number of pixels added at the beginning of axis i and xi_end is the number of pixels added at the end of axis i. If not present, the padding defaults to 0 along start and end of each spatial axis.
Default value “empty”.
strides : array, 1-dimensional tensor with stride value along each spatial axis. If not present, the stride defaults to 1 along each spatial axis.
Default value “empty”.
training? : boolean, whether the layer is in training mode (can store data for backward).
Default value “True”.
lda coeff : float, defines the coefficient by which the loss derivative will be multiplied before being sent to the previous layer (since during the backward run we go backwards).
Default value “1”.

name (optional) : string, name of the node.

Output parameters

output (heterogeneous) – T : object, output tensor produced by rearranging blocks into an image.

Type Constraints

T in (tensor(bfloat16), tensor(bool), tensor(complex128), tensor(complex64), tensor(double), tensor(float), tensor(float16),
tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(string), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)) : Constrain input and output types to all numeric tensor types.

Example

All these exemples are snippets PNG, you can drop these Snippet onto the block diagram and get the depicted code added to your VI (Do not forget to install Deep Learning library to run it).

Deep Learning

Installation guide

Execution providers

General

Architecture

Layers

Nodes

Nodes

Activation

Mono Input

Parameters

Graph Function

Graph

File

Get & Set

Runtime

Create

Academic Training

Inference

Training

Exec

Academic

Input

Advanced

Add Weight

Index

Name

Convert

From ONNX

To ONNX

Format Weight

Get Weight

Metrics

Pre/Post Process

Set Weight

More

Layers parameters

Nodes Parameters

Accelerator

Quick start

Installation guide

Execution providers

General

API

Execution

Open

Graph

File

Exec

Constant

Generator

Full Train Step

Eval Step

Train Step