OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Neural Image Assessment

Neural Image Assessment is a Python and PyTorch implementation of NIMA, a neural model for estimating image aesthetic quality. It includes scripts to train on the AVA dataset and run inference with a saved model.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Neural Image Assessment — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when you need a modifiable, local PyTorch implementation for training or image-aesthetic inference, rather than a hosted scoring service.

Who it is for

  • Developers who want to generate aesthetic-rating predictions for image collections.
  • Researchers or practitioners adapting a PyTorch image-assessment model trained on AVA.

Documented uses

  • Train the model using AVA images and rating annotations.
  • Run inference with a saved model and export prediction files.
  • Use model scores to compare images or support photo-review and collection-analysis tools.
Before you choose

Requirements & limits

  • The documented environment uses Python, PyTorch 1.8.1 and CUDA 11.1, with Conda environment setup provided.
  • Training requires AVA images and annotation CSVs; curated train, validation and test CSVs are linked.
  • The repository links a pretrained model checkpoint for warm-start training.
  • The README cautions that the example training was not fully converged and that predictions are sensitive to contrast, with weaknesses at extreme aesthetic ratings.
  • The repository identifies its license as MIT.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Run inference with a saved NIMA model

Generate prediction files for images using an existing model.

Before you start

  • Python environment with the repository dependencies; the README specifies PyTorch 1.8.1 and CUDA 11.1.
  • A saved model, test annotation CSV and image directory.

Steps

  1. Prepare the test CSV and image directory referenced by its entries.

  2. Run the repository's test.py script with the model, CSV, images and prediction-output paths.

Expected outcome

Predictions are written to the selected output location.

Guide limitations

  • This inference workflow requires an existing model; training is a separate step.
  • The README reports that predictions can be sensitive to contrast and that the model often fails on images with very low or high aesthetic ratings.
Documentation 1 ↗
Source-based practical guide

Train NIMA on the AVA dataset

Train the repository's image-aesthetic assessment model using AVA images and annotation CSVs.

Before you start

  • AVA images downloaded and decompressed into an images/ directory.
  • Train and validation annotation CSVs.
  • The repository's Conda environment file; the README specifies PyTorch 1.8.1 and CUDA 11.1.

Steps

  1. Create the environment using the repository's env.yml file.

  2. Download and decompress the AVA dataset, then obtain the curated annotation CSVs.

  3. Run main.py with the image path, training flag, train and validation CSV paths, checkpoint path and training options shown in the README.

Expected outcome

The training script runs with validation and early stopping settings and saves checkpoints to the specified path.

Guide limitations

  • The author reports that the setup had not fully converged and had limited hyperparameter tuning.
  • The README's example command specifies 100 epochs and early-stopping patience of 10; these are example settings, not a guaranteed training result.
Documentation 1 ↗
COMPARE OPTIONS

Compare options for Model inference and serving

These projects document a common task and share a project role. This is a source-based comparison, not an execution benchmark or a guarantee of interchangeability.

ProjectDocumented purposePlatformAccess and requirements
Neural Image AssessmentTrain the model using AVA images and rating annotations.Cloud, Linux, macOS, Windows
Pulse Of MotionEstimate a video's physical frame rate from its visual dynamics.LinuxThe repository is public. Its README describes a released checkpoint that downloads from Hugging Face on first inference, with a manual download option. The strongest internal model is not fully released; the README directs interested users to contact the project about access. No license information is provided in the repository evidence.
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
18 October 2026
Original website title
yunxiaoshi/Neural-IMage-Assessment · Official homepage title · retained in its source language
How we review sourcesSuggest a correction