Personal project

Image Denoising with CAE & U-Net

A comparison of convolutional autoencoders and U-Net for removing salt-and-pepper noise from PlantVillage tomato leaf images.

Computer Vision & Image RestorationDeep learning project

Explore feature spotlights

Five leaf image examples arranged in rows for salt-and-pepper noise, Median Filter, CAE, CAE-SC, U-Net, and ground truth

Tech Stack

PythonPyTorchCAEU-NetGoogle Colab

Feature spotlights / Image Denoising with CAE & U-Net

Explore the features of this project.

Choose a feature to explore its workflow, screenshots, or demo videos.

Choose another spotlight

Visual restoration results

The visual comparison documents salt-and-pepper noise removal with classical filtering and neural networks. Shared samples help connect the numerical evaluation with the appearance of the restored images.

Visual restoration results

Leaf image restoration

Five leaf samples compare noisy inputs with results from Median Filter, CAE, CAE-SC, and U-Net, followed by the ground-truth images. The aligned rows make remaining noise, smoothing, and retained leaf detail visible across methods.

Compare the restored images with their references, including leaf edges and texture.

Challenge

Noise can obscure visual information used by agricultural computer vision systems.

Restore noisy leaf images while preserving detail and comparing architectures.

Approach

Compared median filtering, CAE, CAE-SC, and U-Net using PyTorch, evaluating restored images with PSNR and SSIM.

Investigates image restoration methods that can support plant disease diagnosis workflows.

Implementation

Convolutional autoencoder
Autoencoder with skip connections
U-Net encoder-decoder
Median filter comparison
PSNR and SSIM image quality evaluation

Features

Salt-and-pepper noise removal
PlantVillage tomato leaf images
Median filter baseline
CAE, CAE-SC, and U-Net comparison
PSNR and SSIM evaluation

Highlights

Comparison of classical filtering and deep learning
Skip connections and encoder-decoder models
Restoration evaluation on agricultural images

Let's talk about your next project.

Let's Talk

Loading page...

Back to projects

Personal project

Image Denoising with CAE & U-Net

A comparison of convolutional autoencoders and U-Net for removing salt-and-pepper noise from PlantVillage tomato leaf images.

Computer Vision & Image RestorationDeep learning project

Explore feature spotlights

Tech Stack

PythonPyTorchCAEU-NetGoogle Colab

Feature spotlights / Image Denoising with CAE & U-Net

Explore the features of this project.

Choose a feature to explore its workflow, screenshots, or demo videos.

Choose another spotlight

Visual restoration results

Leaf image restoration

Compare the restored images with their references, including leaf edges and texture.

Challenge

Noise can obscure visual information used by agricultural computer vision systems.

Restore noisy leaf images while preserving detail and comparing architectures.

Approach

Compared median filtering, CAE, CAE-SC, and U-Net using PyTorch, evaluating restored images with PSNR and SSIM.

Investigates image restoration methods that can support plant disease diagnosis workflows.

Implementation

Convolutional autoencoder
Autoencoder with skip connections
U-Net encoder-decoder
Median filter comparison
PSNR and SSIM image quality evaluation

Features

Salt-and-pepper noise removal
PlantVillage tomato leaf images
Median filter baseline
CAE, CAE-SC, and U-Net comparison
PSNR and SSIM evaluation

Highlights

Comparison of classical filtering and deep learning
Skip connections and encoder-decoder models
Restoration evaluation on agricultural images

Let's talk about your next project.

Let's Talk