Skip to content

In this project, I developed and trained a model that uses the Deep Lab V3 Plus architecture for image segmentation — trained particularly on human figures (faced, bodies, et cetera). The model as well as the code to run the model has been provided.

License

Notifications You must be signed in to change notification settings

ryan-air/Image-Segmentation-DeepLabV3Plus

Repository files navigation

Image-Segmentation-DeepLabV3Plus

Passion Project

Image segmentation is the process of dividing an image into multiple meaningful and distinct regions or objects. It involves assigning a label or class to each pixel or region in an image to differentiate between different objects or areas of interest. This allows for the extraction and analysis of specific objects or regions within an image for various applications such as object recognition, scene understanding, and computer vision tasks. Through the use of TensorFlow, this project uses image segmentation to create masks that allow us to remove the background of an image.

Deep-Lab-V3 Plus Architecture

Training Parameters

  1. batch_size = 2
  2. learning_rate = 1e-4
  3. epochs = 10

Training/Testing Results

Accuracy F1 Jaccard Recall Precision
0.9205 0.8787 0.8092 0.9312 0.8645

About

In this project, I developed and trained a model that uses the Deep Lab V3 Plus architecture for image segmentation — trained particularly on human figures (faced, bodies, et cetera). The model as well as the code to run the model has been provided.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published