While manually annotating an image, i was not able to rotate the square. Wud be helpful if i could rotate the square to capture rotated images
Roboflow currently has bounding box object detection. For rotated bounding boxes, that would be polygons - but if you are referring to augmentations, you can apply the Rotate augmentation.
Object detection models only detect rectangles aligned with an image; this helps them be really fast. And almost always this is good enough to solve the business problem at hand.
When you would need/want polygons are for particularly oblong objects where a box does a really bad job of fitting it (Eg most of the box is noise vs signal) or where you want to measure the area (in the number of pixels) of the detected object.
This blog post outlines the difference between different computer vision project types: Introduction to Computer Vision
*NOTE: Roboflow will soon be releasing support for polygons!