GIACOMO BORACCHI - TEACHING
 

Image Classification: Modern Approaches (PhD Course)  


AA 2017/2018,Milano

Mission:
The goal of this course is to provide students with an understanding of the most important image-classification algorithms, and in particular of the feature-extraction phase, which is often their most critical component. We will provide an overview of both hand-crafted features, that are still adopted in many industrial and automation-control scenarios, and data-driven (i.e. learned) features, which have recently become a standard in challenging natural-image recognition problems with huge training datasets.
More information on the Course program page

Calendar:
  • February 12th, Introduction and basics of image handling in Python, from 14:00 to 18:00 in Aula S.0.2. Ed 3
  • February 14th, Hand-crafted features for image classification, from 14:00 to 18:00 in Aula S.0.2. Ed 3
  • February 16th, Computer Vision features for image classification, from 14:00 to 18:00 in Aula S.0.5. Ed 3
  • February 19th, Data-driven features: Convolutional Neural Networks from 14:00 to 18:00 in Aula S.0.5. Ed 3
  • February 21th, Advanced CNNs and Best practices in image classification, from 14:00 to 18:00 in Aula S.0.2. Ed 3
  • February 23th, An overview on extended problems, from 14:00 to 18:00 in Aula S.0.5. Ed 2

Teaching Materials:

Lez 1: Introduction to IC + Local Spatial Transformation  (Slides 1, Slides 2), (Python Codes)

Lez 2: Hand Crafted Features + Local Spatial Transformation (cnt) + Global Transformations   (Slides 1), (Slides 2), Dataset (now updated to 4 classes and 2.1K samples) for Parcel Classification, (Python Codes)

Lez 3: Computer Vision Features  (Slides) Dataset (now updated to 4 classes and 2.1K samples) for Parcel Classification (Python Codes)

Lez 4-5-6: Convolutional Neural Networks  (All the updated slides) Dataset Rock Paper Scissors

Lez 4: Introduction to Keras  (Python Codes)

Lez 5: RPS examples updated  (Python Codes)

Lez 6: Augmentation + Assignments  (Python Codes), (Video: quadcopter navigation by CNN)

Lez 6: Generative Models   (Slides on GAN)

Lez 6: IC: Research Activities and Thesis Opportunities  (Slides)