Implementation:LaurentMazare Tch rs Imagenet Load Image And Resize224

Knowledge Sources	tch-rs
Domains	Computer_Vision, Data_Preprocessing
Last Updated	2026-02-08 14:00 GMT

Overview

Concrete tool for loading and preprocessing images for ImageNet-pretrained models provided by the tch vision module.

Description

imagenet::load_image_and_resize224 loads an image from the filesystem, resizes it to 224x224 pixels, and applies ImageNet normalization. Internally it delegates to image::load_and_resize (using the vendored stb_image library) and then normalizes each channel using the ImageNet mean and standard deviation. The output is a float32 tensor of shape [3, 224, 224].

Usage

Use this before running inference with any pretrained ImageNet model. The returned tensor should be unsqueezed to add a batch dimension before being passed to the model.

Code Reference

Source Location

• Repository: tch-rs
• File: src/vision/imagenet.rs
• Lines: 46-48

Signature

pub fn load_image_and_resize224<T: AsRef<Path>>(path: T) -> Result<Tensor, TchError>

Import

use tch::vision::imagenet;

I/O Contract

Inputs

Name	Type	Required	Description
path	T: AsRef<Path>	Yes	Path to image file (supports jpg, png, tga, bmp)

Outputs

Name	Type	Description
Result<Tensor>	Tensor	Float32 tensor of shape [3, 224, 224], ImageNet-normalized

Usage Examples

use tch::vision::imagenet;

let image = imagenet::load_image_and_resize224("photo.jpg")?;
println!("{:?}", image.size());  // [3, 224, 224]

// Add batch dimension for model input
let input = image.unsqueeze(0);  // [1, 3, 224, 224]

Related Pages

Implements Principle

• Principle:LaurentMazare_Tch_rs_ImageNet_Preprocessing