4 Advanced transfer learning use-cases for computer vision: part-ii
This chapter covers
- Building an understanding of some of the most popular multi-stage and single-stage object detection models/architectures
- Developing object detection pipeline for custom datasets leveraging pre-trained object detection models
- Understanding image segmentation task in detail along with key improvements based on popular segmentation models
- Exploring hands-on examples using transfer learning for the task of image segmentation
In the last chapter we looked at a object detection workflow along with a few metrics that are typically used to evaluate such models. Object detection researchers come from two schools of thought: multi-stage object detection, which is slow but more performant, and single-stage object detection which is fast at the cost of lower detection performance. We will dive deeper into such aspects in subsequent sections. In the previous chapter, we also briefly touched upon the task of image segmentation as a more fine-grained related task to object detection. While object detection is helpful in locating specific objects in a given image, segmentation helps in identification of exact boundary of such identified objects.