PROJECT TITLE :

Divide and Count Generic Object Counting by Image Divisions

ABSTRACT:

No prior categorization information is required to count the number of objects. There is no need for any local annotations to estimate global image counts. Image divisions are created using our method, each encompassing the entire image. Images are divided into regions or grid cells by region suggestions. To anticipate global picture counts, our solution uses an end-to-end deep learning architecture. An item count is predicted by using a counting layer, which ensures consistency in count values even when dealing with overlapping parts of an image. Set theory's inclusion-exclusion concept informs our counting layer. With the help of Pascal-VOC2007, we dissect each component of our proposed strategy and test its efficacy on a variety of real-world data sets, including the large-scale MS-COCO generic object dataset and three class-specific counting datasets, including pedestrian data from UCSD and car data from CARPK and PUCPR+.