IITM BS Deep Learning for Computer Vision (BSDA5006): Syllabus and Tips

Deep Learning for Computer Vision (DLCV, BSDA5006) is a 4 credit elective at the degree level of the IITM BS in Data Science. The handbook places it at level 5. The DS course page and the DS handbook list no prerequisite. The course starts with classic image processing, then covers CNNs, detection, segmentation, attention and generative models for images.

The course is taught by Prof. Vineeth N B, Professor of Computer Science and Engineering at IIT Hyderabad.

Handbook check. The sources do not fully agree:

• The DS course page and DS course table give no prerequisite. The ES handbook lists this course as an open elective for ES students with Deep Learning (CS3004) as the prerequisite.
• The DS minor list writes the code as BSCS5006. The ES handbook uses DA5006 in one table and CS5003 in another.

The name is the same everywhere. Check the code and prerequisite shown when you register.

What you learn in Deep Learning for Computer Vision

The 12 weeks fall into four blocks.

• Weeks 1 to 3: classic computer vision. How images are formed and stored, filtering and convolution. Then finding edges, corners and blobs, and features such as SIFT, SURF and HoG. Then matching images with bag-of-words, RANSAC, the Hough transform and optical flow.
• Weeks 4 to 7: CNNs and what they can do. A review of deep learning and backpropagation. Then CNN designs from AlexNet to ResNet and DenseNet, and ways to see inside a CNN such as Grad-CAM and style transfer. Then CNNs for recognition and verification with Siamese networks, object detection (R-CNN family, YOLO, SSD) and segmentation (U-Net, Mask R-CNN).
• Weeks 8 and 9: video, attention and language. RNNs combined with CNNs for video and action recognition. Then attention in vision, image captioning, visual question answering and transformers.
• Weeks 10 to 12: generating images and recent trends. GANs, VAEs and other generative models, and their uses such as inpainting and super-resolution. Then zero-shot and few-shot learning, self-supervised learning and reinforcement learning in vision.

The suggested books include Deep Learning by Goodfellow, Bengio and Courville, and Computer Vision: Algorithms and Applications by Szeliski.

How the course is assessed

The course page points to the standard course structure on the programme's Academics page. It does not list a separate assessment plan.

Where it counts

• Minor in Multimodal AI Systems, with Speech Technology (BSEE4001) and Large Language Models (BSDA5004). The minor also needs Deep Learning Practice (BSDA5013) as its prerequisite. See the Multimodal AI Systems minor.
• BS level stream rule. It is level 5 and tagged BD, which appears to mean the data science stream. See the BS degree level.
• Terms. The DS course table (updated 18 March 2026) marks it for May 2026, September 2026 and January 2027.

Who finds it hard and how to prepare

The course is very wide. Weeks 1 to 3 are classic image processing with its own maths. Weeks 5 to 7 then pack many architectures into three weeks. Week 4 reviews deep learning in a single week, which is not much time to catch up if you have not done the Deep Learning course.

• Take Deep Learning (BSCS3004) first, even though the DS page does not require it. It covers CNNs and the classic architectures in detail.
• Revise 2D convolution by hand on a small grid of numbers. Weeks 1 and 5 both rely on it.
• Install OpenCV and try edge and corner detection on your own photos during weeks 1 to 3.
• For detection models, draw a simple diagram of each (R-CNN, Faster R-CNN, YOLO) and note what changed from the previous one.
• Run a pretrained model with Grad-CAM on a few images. It makes the visualisation week much clearer.

What to take before and after

Deep Learning is the best base. For the Multimodal AI minor, add Deep Learning Practice (BSDA5013), whose last four weeks are hands-on vision tasks such as detection and super-resolution. Mathematical Foundations of Generative AI goes deeper into GANs, VAEs and diffusion.

Common questions

Can ES students take this course?

The ES handbook lists it as an open elective for ES students, with Deep Learning (CS3004) as the prerequisite. ES students should confirm with their own support team before registering.

Is DLCV mostly theory or coding?

The page lists both: knowledge goals for each topic and hands-on goals for image processing, deep vision models and advanced tasks. It does not say how much of the grade comes from coding.

How is DLCV different from Deep Learning Practice?

DLCV covers computer vision in depth over all 12 weeks. Deep Learning Practice spends four weeks each on text, speech and vision, and focuses on training and deploying models in practice.