# deeplab-utils

**Repository Path**: mirrors_DiUS/deeplab-utils

## Basic Information

- **Project Name**: deeplab-utils
- **Description**: Utilities for working with DeepLab
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2021-10-22
- **Last Updated**: 2026-03-08

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# deeplab-utils
Utilities for working with DeepLab, a neural network for image segmentation.

This is a collection of tools for working with the DeepLab deep learning model, as described in the paper [Rethinking Atrous Convolution for Semantic Image Segmentation](https://arxiv.org/pdf/1706.05587.pdf). We used the model implementation [here](https://github.com/rishizek/tensorflow-deeplab-v3).

In the first instance we needed to create our own masks for training. We used [Labelbox](labelbox.io) for segmenting our images, and exported the project in their JSON format. Within this file there is a reference to a binary mask for each label category. However, the DeepLab implementation assumes the training set to be in Pascal VOC format, which combines the category masks into a combined, single-channel instance mask. Each pixel in the combined mask has a value corresponding to the category number.