OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Image & Video AI

Deconvfaces

Deconvfaces is an open-source Python command-line application for training deconvolution networks on face datasets and generating face images, random sequences, or interpolations between keyframes.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Deconvfaces — OpenViST project profile in Image & Video AI.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Deconvfaces when you can provide a face dataset and want to train or run a deconvolution model through a Python CLI, with YAML parameters controlling image generation.

Who it is for

  • Developers and researchers seeking a scriptable workflow for training face-generation models.
  • Creators who want to generate varied face images or keyframe-based face sequences from a trained model.

Documented uses

  • Train a deconvolution network on a face dataset.
  • Generate single or random face images from a trained model.
  • Create a more contiguous sequence of generated images.
  • Interpolate between specified keyframes to create an animation.
Before you choose

Requirements & limits

The repository is public and identifies the project as MIT-licensed.

  • Requires Python 3 and Keras, NumPy, SciPy and tqdm.
  • GPU memory affects the feasible number of deconvolution layers and image size.
  • The repository includes example YAML parameter files for generation modes.
  • The README notes that the model was unable to learn interpolation between orientations.

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

Train a face-generation model

Start model training on a face dataset using the documented CLI workflow.

Before you start

  • Python 3
  • Keras, NumPy, SciPy and tqdm
  • A face dataset and a GPU with sufficient memory for the chosen model settings

Steps

  1. Provide the dataset path to the training command: the command below.

    Documented command
    python3 faces.py train path/to/data

  2. Optionally adjust the number of deconvolution layers with -d, batch size with -b, or kernels per layer with -k to suit the available memory.

Expected outcome

The application begins training a deconvolution model on the provided data.

Guide limitations

  • More deconvolution layers require enough GPU memory to support larger images.
  • Adjusting batch size or kernel count to fit memory may result in worse results or longer training.
Documentation 1 ↗
Source-based practical guide

Generate images or an interpolated sequence

Use a trained model and YAML parameters to produce face images or an animation between keyframes.

Before you start

  • Python 3
  • Keras, NumPy, SciPy and tqdm
  • A trained model and a YAML parameter file

Steps

  1. Choose or prepare a YAML parameter file; example files are available in the repository's params directory.

  2. Run the generation command with the model, output directory and parameter file: the command below.

    Documented command
    python3 faces.py generate -m path/to/model -o output/directory -f path/to/params.yaml

  3. Select single, random, drunk, or interpolate mode in the parameters for a single image, random images, a more contiguous sequence, or interpolation between specified keyframes.

Expected outcome

The selected mode generates images in the specified output directory, or an animation sequence between keyframes.

Guide limitations

  • The README reports that its model was unable to learn interpolation between orientations.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
5 November 2026
Original website title
flynnmd/deconvfaces · Official homepage title · retained in its source language
How we review sourcesSuggest a correction