sujitpal commited on
Commit
9cde513
1 Parent(s): 260b8b7

fix: add explanation

Browse files
dashboard_image2image.py CHANGED
@@ -56,13 +56,15 @@ def app():
56
  (10k images and ~50k captions from the remote sensing domain).
57
 
58
  This demo shows the image to image retrieval capabilities of this model, i.e.,
59
- given an image file name as a query (we suggest copy pasting the file name
60
- from the result of a text to image query), we use our fine-tuned CLIP model
61
  to project the query image to the image/caption embedding space and search
62
  for nearby images (by cosine similarity) in this space.
63
 
64
  Our fine-tuned CLIP model was previously used to generate image vectors for
65
  our demo, and NMSLib was used for fast vector access.
 
 
 
66
  """)
67
 
68
  image_file = st.text_input("Image Query (filename):")
 
56
  (10k images and ~50k captions from the remote sensing domain).
57
 
58
  This demo shows the image to image retrieval capabilities of this model, i.e.,
59
+ given an image file name as a query, we use our fine-tuned CLIP model
 
60
  to project the query image to the image/caption embedding space and search
61
  for nearby images (by cosine similarity) in this space.
62
 
63
  Our fine-tuned CLIP model was previously used to generate image vectors for
64
  our demo, and NMSLib was used for fast vector access.
65
+
66
+ You will need an image file name to start, we recommend copy pasting the
67
+ file name from one of the results of the text to image search.
68
  """)
69
 
70
  image_file = st.text_input("Image Query (filename):")
dashboard_text2image.py CHANGED
@@ -61,6 +61,9 @@ def app():
61
 
62
  Our fine-tuned CLIP model was previously used to generate image vectors for
63
  our demo, and NMSLib was used for fast vector access.
 
 
 
64
  """)
65
 
66
  query = st.text_input("Text Query:")
 
61
 
62
  Our fine-tuned CLIP model was previously used to generate image vectors for
63
  our demo, and NMSLib was used for fast vector access.
64
+
65
+ Some suggested queries to start you off with -- "ships", "school house",
66
+ "military installations", "mountains", "beaches", "airports", "lakes", etc.
67
  """)
68
 
69
  query = st.text_input("Text Query:")