I see that in fastai 2019 course part2, there is no object detection course anymore. The subject is covered (I assume) in part 1 with the segmentation.
So if I want to perform object detection, I shall rely on 2018 version of the course? Also just for understanding, why has this change been made ?
Part 2 is a foundational course now, and shows how fastai v2 will look. Jeremy has mentioned inklings of a object detection course soon, but with v2 dev full time do not expect this quickly. For now, we have documentation on how to use the API, and the 2018 lecture:
One idea to consider is to look at RepPoints which was just updated last week and appears to have SOTA results. No bounding boxes used! And they have full code released on github.
I’m hoping to have a chance to work with it and integrate it into FastAI framework. Here’s their paper for now:
Modern object detectors rely heavily on rectangular bounding boxes, such as
anchors, proposals and the final predictions, to represent objects at various
recognition stages. The bounding box is convenient to use but provides only a
Code is linked there.
hey. can see here
We’re you able to integrated with FastAi ?
Hey! just checking if there is any latest object detection course/video that the fastai course team released? (or) should I look over the 2018 version only yet?