I found out that you used the official code to compute the SOA scores.
But in[OP-GAN](https://github.com/ppjh8263/semantic-object-accuracy-for-generative-text-to-image-synthesis/tree/1d07bf250aedf9e1b0c55505eb76c49d60ce0055/SOA. ") it is described as : "In order to calculate the IoU scores you need to save the "ground truth" information, i.e. the bounding boxes you give your model as input, so we can compare them with the bounding boxes from the detection network." And there is no bounding box input in your model(xmcgan).
Can I get the details of your SOA score calculation？ Thank you.
I found out that you used the official code to compute the SOA scores. But in[OP-GAN](https://github.com/ppjh8263/semantic-object-accuracy-for-generative-text-to-image-synthesis/tree/1d07bf250aedf9e1b0c55505eb76c49d60ce0055/SOA. ") it is described as : "In order to calculate the IoU scores you need to save the "ground truth" information, i.e. the bounding boxes you give your model as input, so we can compare them with the bounding boxes from the detection network." And there is no bounding box input in your model(xmcgan). Can I get the details of your SOA score calculation？ Thank you.