Regarding the resize of heatmap

[xiaopengguo]

Harmonization/harmonization/common/clickme_dataset.py

Lines 53 to 55 in ffe1c49

	heatmap = tf.image.resize(heatmap, (64, 64), method="bilinear")
	heatmap = gaussian_blur(heatmap, GAUSSIAN_KERNEL)
	heatmap = tf.image.resize(heatmap, (224, 224), method="bilinear")

I'm curious about why apply the Gaussian blur to a smaller heatmap, and then resize it to a larger heatmap. Dose this process result in potential loss of details and is it possible to apply Gaussian blur directly to the larger heatmap? Looking forward to further insights, and thank you in advance!