Silvia Terragni commited on
Commit
1977c6a
β€’
1 Parent(s): aacfe19

rephrasing

Browse files
Files changed (1) hide show
  1. localization.py +5 -4
localization.py CHANGED
@@ -112,9 +112,10 @@ def app():
112
  how the similarity to the image description changes. If you want to have a look at the implementation in details
113
  you can find it in [this Colab](https://colab.research.google.com/drive/10neENr1DEAFq_GzsLqBDo0gZ50hOhkOr?usp=sharing).
114
 
115
- On the two parameters: the pixel size defines the resolution of the localization map. A pixel size of 15 means
116
- that 15 pixels in the original image will form 1 pixel in the heatmap. The refinement
117
- iterations are just a cheap operation to reduce background noise. Too few iterations will leave a lot of noise.
 
118
  Too many will shrink the heatmap too much.
119
 
120
 
@@ -162,7 +163,7 @@ def app():
162
  st.error("Please choose one image and at least one label")
163
  else:
164
  with st.spinner("Computing... This might take up to a few minutes depending on the current load πŸ˜• \n"
165
- "[Colab Link](https://colab.research.google.com/drive/10neENr1DEAFq_GzsLqBDo0gZ50hOhkOr?usp=sharing)"):
166
  heatmap, image = get_heatmap(image_url, caption, pixel_size, iterations)
167
 
168
  with col1:
 
112
  how the similarity to the image description changes. If you want to have a look at the implementation in details
113
  you can find it in [this Colab](https://colab.research.google.com/drive/10neENr1DEAFq_GzsLqBDo0gZ50hOhkOr?usp=sharing).
114
 
115
+ On the two parameters:
116
+ + the *pixel size* defines the resolution of the localization map. A pixel size of 15 means
117
+ that 15 pixels in the original image will form 1 pixel in the heatmap.
118
+ + The *refinement iterations* are just a cheap operation to reduce background noise. Too few iterations will leave a lot of noise.
119
  Too many will shrink the heatmap too much.
120
 
121
 
 
163
  st.error("Please choose one image and at least one label")
164
  else:
165
  with st.spinner("Computing... This might take up to a few minutes depending on the current load πŸ˜• \n"
166
+ "Otherwise, you can use this [Colab notebook](https://colab.research.google.com/drive/10neENr1DEAFq_GzsLqBDo0gZ50hOhkOr?usp=sharing)"):
167
  heatmap, image = get_heatmap(image_url, caption, pixel_size, iterations)
168
 
169
  with col1: