# ROI calculation

In this chapter, I am confused to ROI calculation, then I search from google, this web is helpful:

Region of interest pooling explained (deepsense.ai)

The calculation process as below:

1. A single 8×8 feature map, one region of interest and an output size of 2×2.

2. The region proposal (top left, bottom right coordinates): (0, 3), (7, 8).

3. By dividing it into (2×2) sections (because the output size is 2×2). int((7+0)/2)=3 and int((3+8)/2)=5, which split the region proposal into 4 part.

4. Select the max values in each of the sections.

# interpretation of this chapter’s ROI calculation

To this chapter, why the `rois = torch.Tensor([[0, 0, 0, 20, 20], [0, 0, 10, 30, 30]])` of

``````
tensor([[[[ 0.,  1.,  2.,  3.],

[ 4.,  5.,  6.,  7.],

[ 8.,  9., 10., 11.],

[12., 13., 14., 15.]]]])

``````

is

``````
tensor([[[[ 5.,  6.],

[ 9., 10.]]],

[[[ 9., 11.],

[13., 15.]]]])

``````

To calculate `[0, 0, 10, 30, 30]` ROI, we have to spilt element of matrix from (0,1) to (3,3) into four part. (which 1=10*0.1, 0.1 is the torchvision.ops.roi_pool(X, rois, output_size=(2, 2), spatial_scale=0.1)'s spatial_scale).

Calculation: int((0+3)/2)=1, int((1+3)/2)=2.

Result like image below:

Select max from each part is

``````
[ 9., 11.],

[13., 15.]

``````