Deep Learning For Visual Computing Course | IIT Kharagpur | Prof. Debdoot Sheet
Course Details
| Exam Registration | 258 |
|---|---|
| Course Status | Ongoing |
| Course Type | Elective |
| Language | English |
| Duration | 12 weeks |
| Categories | Electrical, Electronics and Communications Engineering |
| Credit Points | 3 |
| Level | Postgraduate |
| Start Date | 19 Jan 2026 |
| End Date | 10 Apr 2026 |
| Enrollment Ends | 02 Feb 2026 |
| Exam Registration Ends | 20 Feb 2026 |
| Exam Date | 25 Apr 2026 IST |
| NCrF Level | 4.5 — 8.0 |
Master the Future of AI with Deep Learning for Visual Computing
In the rapidly evolving field of artificial intelligence, the ability to interpret and understand visual data is paramount. From medical diagnostics to autonomous vehicles, deep learning has revolutionized how machines perceive the world. If you're a postgraduate student or professional in Electrical, Electronics, or Computer Sciences looking to master these cutting-edge skills, a new course from the Indian Institute of Technology Kharagpur offers a definitive pathway.
About the Course Instructor: Prof. Debdoot Sheet
This 12-week intensive course is led by Prof. Debdoot Sheet, an esteemed Assistant Professor in the Department of Electrical Engineering at IIT Kharagpur. Prof. Sheet is also the founder of SkinCurate Research and brings a wealth of expertise to the classroom.
His academic credentials include MS and PhD degrees in computational medical imaging and machine learning from IIT Kharagpur, complemented by experience as a DAAD visiting PhD scholar at TU Munich. His research spans deep learning, domain adaptation, computational medical imaging, and surgical analytics. A prolific contributor to top-tier journals and conferences, Prof. Sheet is also an active member of IEEE, SPIE, ACM, and serves as an Editor for IEEE Pulse.
What is Deep Learning for Visual Computing?
Deep learning is a subset of machine learning that uses layered, hierarchical architectures to learn abstractions from data. In visual computing, this means teaching a machine to understand an image in strata—much like how humans do.
Consider recognizing a person standing before a mountain in a photo. A deep learning model doesn't see the scene holistically at first. Instead:
- Lower layers identify basic features: edges, lines, colors, and curves.
- Middle layers combine these to recognize shapes, textures, and object parts (like a limb or a tree).
- Higher layers synthesize this information to identify complex objects (a human, a mountain).
- The final layer makes the contextual judgment: "Mr. X is standing in front of Mt. E."
This course is designed to unravel this "hierarchical logic," teaching you not just the theory but also how machines can learn these representative features autonomously. You'll explore applications ranging from image captioning and synthetic image generation to the technology behind self-driving cars.
Course Details at a Glance
| Aspect | Detail |
|---|---|
| Duration | 12 Weeks |
| Level | Postgraduate |
| Primary Categories | Electrical, Electronics and Communications Engineering, Computer Sciences |
| Prerequisites | Knowledge of Digital Image Processing & Machine Learning |
| Tools & Languages | Python, PyTorch |
| Industry Support | Intel, Microsoft, Google, NVIDIA, Philips, Siemens, Tesla, TCS, Infosys, and many more. |
Weekly Course Layout: A Journey from Fundamentals to Advanced Models
The course is meticulously structured to build your expertise from the ground up over 12 weeks.
Weeks 1-4: Foundations of Neural Networks
- Week 1: Introduction to Visual Computing and Neural Networks.
- Week 2: Evolving from Multilayer Perceptrons to Deep Neural Networks with Autoencoders.
- Weeks 3 & 4: Deep dive into Autoencoders for representation learning, initialization, and advanced variants like Stacked, Sparse, and Denoising Autoencoders.
Weeks 5-9: Mastering Convolutional Neural Networks (CNNs)
- Week 5: Core concepts of Cost functions, Learning Rate Dynamics, and Optimization.
- Week 6: Introduction to CNNs and the pioneering LeNet architecture.
- Week 7: Convolutional Autoencoders and deeper architectures like AlexNet and VGGNet.
- Week 8: Exploring very deep networks: GoogLeNet, ResNet, and DenseNet.
- Week 9: Tackling Computational Complexity and the power of Transfer Learning.
Weeks 10-12: Advanced Applications and Models
- Week 10: Moving beyond classification to Object Localization (R-CNN) and Semantic Segmentation.
- Week 11: Unleashing creativity with Generative Models and Adversarial Learning (GANs).
- Week 12: Handling sequential visual data with Recurrent Neural Networks (RNNs) for Video Classification.
Who Should Enroll?
This course is ideally suited for:
- Postgraduate students in Electrical, Electronics, Computer Science, or related fields.
- Professionals and researchers aiming to build or enhance their expertise in deep learning for computer vision.
- Individuals with a solid foundation in Digital Image Processing and basic Machine Learning concepts.
Key Learning Outcomes and Resources
By the end of this course, you will gain both theoretical insights and hands-on coding practice with modern frameworks like PyTorch. The curated exercises are designed to solidify your understanding of current developments in the field.
The course references seminal texts to guide your learning:
- Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep Learning. MIT Press.
- Haykin, S. (2008). Neural Networks and Learning Machines (3rd ed.). Pearson.
With strong backing from leading global industries, the skills you acquire will be directly relevant to the job market, opening doors to roles in tech giants, healthcare imaging, automotive AI, and cutting-edge research.
Embark on this 12-week journey to demystify deep learning and become proficient in building intelligent systems that see and understand our visual world. Enroll today and transform your potential into expertise.
Enroll Now →