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

Gscnn

Gscnn is the official Python implementation of Gated-SCNN, a convolutional neural network for semantic segmentation that incorporates shape information.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Gscnn — 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 Gscnn when you need the official implementation and documented checkpoint-based Cityscapes evaluation workflow for Gated-SCNN.

Who it is for

  • Computer-vision researchers reproducing or evaluating Gated-SCNN.
  • Developers adapting checkpoint-based semantic segmentation in Python.

Documented uses

  • Evaluate a pretrained Gated-SCNN checkpoint on Cityscapes.
  • Use the implementation as a basis for a semantic-segmentation research workflow.
Before you choose

Requirements & limits

  • The documented stack requires Python 3, PyTorch 1.1.0 or later, torchvision and the listed supporting packages.
  • The pretrained model is downloaded separately and must be saved in the checkpoints directory.
  • The README reports training on eight NVIDIA GPUs and warns about single-GPU training with WiderResNet38.
  • The stated CC BY-NC-SA 4.0 license permits non-commercial use subject to its conditions.

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

Evaluate GSCNN on Cityscapes

Run the repository's documented evaluation command with its pretrained Cityscapes checkpoint.

Before you start

  • Clone the GSCNN repository.
  • Use Python 3 and install the listed requirements, including PyTorch 1.1.0 or later and torchvision.
  • Download the pretrained model separately and save it in the checkpoints directory.

Steps

  1. Clone the repository with git clone https://github.com/nv-tlabs/GSCNN and enter the GSCNN directory.

  2. Install the documented Python requirements.

  3. Place the downloaded checkpoint at checkpoints/best_cityscapes_checkpoint.pth.

  4. Run python train.py --evaluate --snapshot checkpoints/best_cityscapes_checkpoint.pth.

Expected outcome

The repository's Cityscapes evaluation workflow is invoked using the specified checkpoint.

Guide limitations

  • The checkpoint must be obtained separately.
  • The README cautions that training with WiderResNet38 may be an issue on a single GPU; this note concerns training, not the documented evaluation command.
Documentation 1 ↗
Evidence and transparency

Sources & review

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