Designing an Image Segmentation Model

An Image Segmentation Model designed to be used within the Computer Vision System of a Self Driving Vehicle.

It is trained based on the Cityscrapes Dataset.

This model identifies different classes of objects in photos captured by a Vehicle's sensors :

• Constructions
• Nature
• Sky
• People
• Vehicle
• Object

One of the challenge of this project was to reduce the computing power needed to train and deploy the model so it could be easily used by an edge device like the self-driving vehicle.

We used transfer learning and managed to reach state-of-the art performance on the Cityscrapes dataset :

The prediction API is then published on a web interface using Flask.

Useful Links

• Project presentation (Powerpoint)
• Technical Report of Model development (Word)
• Jupyter Notebook (Model training)
• Flask Deployment Folder

Screenshots

Encoder-Decoder Framework

Transfer Learning

Final Model Architecture

Web Interface

Model Prediction

Predicts stock images

Libraries / Packages Used

• Tensorflow - Keras
• Flask
• Scikit-Image
• Albumentations
• Open CV
• Segmentation Models
• SqueezeNet Keras Implementation
• Bootstrap
• Matplotlib / Seaborn
• Pandas / Numpy

Developed By

Octave Antoni

Connect with me on Linkedin

License

Copyright 2023 Octave Antoni

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.