Syllabus
Lab Content: (MATLAB/Python/other open-source software) 45 hours
- Image Classification: Training a CNN to classify road signs and traffic
- Object Detection: Implementing a YOLO (You Only Look Once) model for detecting vehicles and
- Lane Detection: Developing a deep learning model to detect and track lane
- Semantic Segmentation: Using a CNN to segment road scenes into different classes (road, vehicles, pedestrians).
- Trajectory Prediction: Implementing an LSTM network to predict the future trajectories of other
- Path Planning: Developing a reinforcement learning algorithm to navigate an autonomous vehicle through a simulated environment.
- Localization: Implementing a particle filter-based localization algorithm to estimate the vehicle’s
- Sensor Fusion: Integrating data from multiple sensors (camera, LiDAR, radar) to detect and track
- Simulation: Building a simulation environment for testing autonomous vehicle algorithms and training
- Real-world Testing: Deploying trained models on an autonomous vehicle platform for real-world testing and
Unit 1
Machine learning – Basic motivation, examples of machine learning applications, supervised, unsupervised and reinforcement learning. Support Vector classification and K-Means clustering. Fundamentals of artificial neural networks (ANNs), Building blocks of neural networks: neurons, layers, and activation functions, Training neural networks using gradient descent and back propagation. ANN based regression model.
Unit 2
Deep learning – Introduction, Convolutional Neural Networks (CNNs) – Understanding CNN architecture, Convolutional layers, pooling layers, and fully connected layers, Training CNNs for image classification tasks. Recurrent Neural Networks (RNNs) -Introduction to RNNs and their applications in sequential data analysis, Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) architectures, Training RNNs for time series prediction.
Unit 3
Advanced Deep Learning Techniques – Introduction to advanced deep learning architectures: Generative Adversarial Network,(GANs), autoencoders, and reinforcement learning, Applications of GANs in generating synthetic data for automotive engineering tasks, Implementing autoencoders for anomaly detection and dimensionality reduction.
Objectives and Outcomes
Course Objectives
- Inculcate the knowledge about various deep learning methods and its automotive applications
- Impart the concepts to formulate a deep learning model for autonomous vehicle
- Enable to use the computational tools for solving real time problems in autonomous
Course Outcomes
CO |
CO Description |
CO1 |
Develop a machine learning model with the help of classification and regressions methods. |
CO2 |
Acquire comprehensive understanding of neural network architectures, optimization algorithms and
activation functions involved in neural networks |
CO3 |
Formulate a deep learning model using various deep learning methods for autonomous vehicle
applications. |
CO4 |
Apply deep learning techniques to solve problems pertinent to autonomous vehicles using
computational tools. |
CO-PO Mapping
|
PO1 |
PO2 |
PO3 |
PO4 |
PO5 |
CO1 |
3 |
2 |
2 |
1 |
2 |
CO2 |
3 |
2 |
1 |
2 |
2 |
CO3 |
3 |
2 |
1 |
2 |
3 |
CO4 |
3 |
2 |
2 |
3 |
3 |
Skills acquired
Develop expertise in autonomous driving technologies, mastering perception algorithms, deep learning, and reinforcement learning for effective implementation in autonomous systems.