Regarding hint area generator.

[shubhamk-76]

Sir, In the architecture of "Hint Area Generator" you have show two arrow going inside the encoder.
Have i understood it right?
Red arrow is Image and Green is the hint(square box).
sir, could you also upload the code of this part, I mean code to generate P using square box.
Looking forewared to hear from you.
Thanks and Regards

[alreadysleep]

After reading the paper, I feel that this P is generated in the following way (because I asked the author whether the warm-up training is simply running the train.py file, and he said yes): 1. Run the train.py file (with the epoch changed to 15), which generates the weight file model_best.pth. 2. Then, use this model_best.pth to run test.py to obtain P.
Additionally, the red arrow in the diagram is labeled as "inference," which I believe should not be simplistically interpreted as inputting an image. Instead, it should be understood as running test.py (I -> P) and running Point2gt.py (P -> S). The green arrow represents the warm-up training, which is running the train.py file, and should not be simplistically interpreted as inputting bounding box prompts (because the train.py file requires input images and labels, where the labels correspond to S in the framework diagram).