Join us for our 9th PyData Montreal meetup:
- "Weakly Supervised Methods for Object Detection and Localization" by Akhil P M
- "Interactive image processing with scikit-image and Dash" by Emmanuelle Gouillart
6:00 pm—Doors open
6:20 pm—Weakly Supervised Methods for Object Detection and Localization
Object detection in the fully supervised settings requires bounding box annotations for each instance present in the image. Such annotations are expensive to obtain and subjective to the human annotators. Due to this reason, weakly supervised methods using only the image-level labels for localization is gaining popularity in the vision community. In this talk, we will discuss weakly supervised methods for object detection and localization. In particular, we will discuss the problem definition, challenges in this reduced supervision settings and some popular weakly supervised algorithms.
Akhil is currently pursuing his Ph.D. in Weakly Supervised methods for object detection and localization with deep architectures. Prior to that, he worked in a telecom firm as Data Scientist where he was developing NLP solutions for customer feedback analytics. His research interests are in unsupervised learning, weakly supervised learning and generative models.
7:00 pm — Break
7:20 pm —Interactive image processing with scikit-image and Dash
Images are an important class of data in science or business. Tasks such as quantification of organ geometry in medical imaging, or construction of training sets and pipelines for machine learning models, typically rely on a combination of interactive user annotations and image processing algorithms. In this talk I will present two open-source packages for interactive image processing from the Python data science ecosystem.
scikit-image is a popular library for processing 2D and 3D images as Numpy numerical arrays, with a focus on scientific imaging and pedagogical example-based documentation. I will show how to use scikit-image for various image processing tasks, from basic preprocessing (eg normalizing image geometry or exposure) to advanced segmentation tasks, and I will discuss future developments planned for scikit-image.
Dash, developed by plotly, is an open-source framework for building interactive analytical web applications in pure Python or R. After a brief demo of how to build an application with Dash, I will present a new component library for Dash called dash-canvas, for interactive image annotation and processing. I will show the current annotation capabilities of dash-canvas, how to integrate annotations with dash-canvas into image processing pipelines, and discuss the roadmap of this young package.
Emmanuelle Gouillart is a scientific Python developer at Plotly (Montreal), where she is a core developer of the `plotly.py` visualization library, and of the more recent `dash-canvas` image annotation library. She has a background in physics and materials science, and she has carried on simultaneously scientific research and software development during the last years. She became a core contributor of Python's popular image processing library scikit-image since a large part of her research relies on extracting quantitative data from image datasets. She has been a co-organizer of the Euroscipy conference during the last ten years, and she enjoys very much discussing with Python users about image processing and visualization at meetups and conferences.
8:00 pm — Break and networking
8:45 pm — Head out for drinks @ Les Soeurs Grises: https://goo.gl/maps/hoZjbNKeBzHHPL8cA