Weakly Supervised Learning for Object Localization

Supervised learning techniques require large quantities of high-quality annotations as labels. For tasks like segmentation, the time and cost of creating annotations are often higher than tasks like classification. To overcome these limitations, one low-cost option is to make use of lower quality annotations collected in large quantities, i.e using weakly supervised learning.

In this article, we’re going to focus on applying WSL techniques to image-based data, to be precise, WSL for object Localization.

Object localization refers to identifying the location of one or more objects in an image and drawing a bounding box around their extent. Object detection combines the localization of an image with the classification of one or more objects in an image.

Object localization locates the presence of objects in an image and highlights their location with a bounding box.

Some top-performing DL models for object detection:

Standard Supervised object detection Vs. Weakly-Supervised Approach

Now, we can use a simple thresholding method to segment the CAM and obtain a bounding box. Firstly, regions in the CAM whose value is above 20% of the max-value of the CAM are used to create a segmentation map. Then, a bounding box is drawn to tightly encompass the largest connected component in the segmentation map.

To read more on WSOL, the following are some papers based on it:

Reach out to me at:

Related posts:

Arranging your advanced education advertising systems are tied in with working around two things: the increasing expense of educational cost and the developing number of understudies applying to…

Project Portfolio Management — or PPM — is the centralized management of the processes, methods, and technologies used by project managers and project management offices to analyze and collectively…

When I distanced myself from writing, to focus in my studies, I prioritized my degrees over my desires. This led to me loosing my audience and making me depressed. I opened my heart through writing…