Closed trouble-maker007 closed 3 weeks ago
null
img_sal = Image.open(self.sal[idx]).convert("L") img_sal_sub = Image.open(self.sal_sub[idx]).convert("L") img_sal_map = Image.fromarray(np.maximum(np.array(img_sal), np.array(img_sal_sub)))
@yuli0103 I could not find any code about how to generate the image about the
saliency_sub
, would you like to give more detail about how to preprocess? Does use two model(ISNet and BASNet) to generate saliency detection result I was tried to preprocess the cgl data into your training format, but still have no idea to get thesaliency_sub
data.![]()
I apologize for the inconvenience. In our setup, images in the 'saliency' directory are generated using the ISNet model, while images in the 'saliency_sub' directory are generated using the BASNet model. You can find the specific code in data_process/saliency_detection.py
. Furthermore, I have updated the latest code package and dataset link.
@yuli0103 Thank you for your reply.I will try to train under this setting.
@yuli0103 Does run single image support condition layout generate, such as cwh
@yuli0103 Does run single image support condition layout generate, such as cwh
Thank you for your suggestion. Going forward, I will support condition layout generation in "run_single_image.py".
@yuli0103 thanks for the response, I set batch 1 in test.py
img_sal = Image.open(self.sal[idx]).convert("L") img_sal_sub = Image.open(self.sal_sub[idx]).convert("L") img_sal_map = Image.fromarray(np.maximum(np.array(img_sal), np.array(img_sal_sub)))
@yuli0103 I could not find any code about how to generate the image about the
saliency_sub
, would you like to give more detail about how to preprocess? Does use two model(ISNet and BASNet) to generate saliency detection result I was tried to preprocess the cgl data into your training format, but still have no idea to get thesaliency_sub
data.