Spaces:
Sleeping
Sleeping
Ana Sanchez
commited on
Commit
·
fa8b5a5
1
Parent(s):
d87e105
update app
Browse files
app.py
CHANGED
|
@@ -305,7 +305,7 @@ def reshape_image(arr):
|
|
| 305 |
##### STREAMLIT FUNCTIONS ######
|
| 306 |
st.title('CLOOME. Bioimage database retrieval from chemical structures (and viceversa)')
|
| 307 |
|
| 308 |
-
def main_page(top_n):
|
| 309 |
st.markdown(
|
| 310 |
"""
|
| 311 |
Contrastive learning for self-supervised representation learning has brought a strong improvement to many application areas, such as computer vision and natural language processing.
|
|
@@ -358,7 +358,7 @@ def molecules_from_image(top_n, model_path):
|
|
| 358 |
molpath = os.path.join(datapath, "mols.hdf")
|
| 359 |
fps_fname = save_hdf(morgan, molnames, molpath)
|
| 360 |
mol_imgs = draw_molecules(smiles)
|
| 361 |
-
mol_features, mol_ids = main(mol_index,
|
| 362 |
predefined_features = False
|
| 363 |
else:
|
| 364 |
mol_index = pd.read_csv(mol_index_file)
|
|
@@ -375,9 +375,11 @@ def molecules_from_image(top_n, model_path):
|
|
| 375 |
print(img_features.shape)
|
| 376 |
print(mol_features.shape)
|
| 377 |
|
|
|
|
|
|
|
| 378 |
logits = img_features @ mol_features.T
|
| 379 |
mol_probs = (30.0 * logits).softmax(dim=-1)
|
| 380 |
-
top_probs, top_labels = mol_probs.cpu().topk(
|
| 381 |
|
| 382 |
# Delete this if want to allow retrieval for multiple images
|
| 383 |
top_probs = torch.flatten(top_probs)
|
|
|
|
| 305 |
##### STREAMLIT FUNCTIONS ######
|
| 306 |
st.title('CLOOME. Bioimage database retrieval from chemical structures (and viceversa)')
|
| 307 |
|
| 308 |
+
def main_page(top_n, model_path):
|
| 309 |
st.markdown(
|
| 310 |
"""
|
| 311 |
Contrastive learning for self-supervised representation learning has brought a strong improvement to many application areas, such as computer vision and natural language processing.
|
|
|
|
| 358 |
molpath = os.path.join(datapath, "mols.hdf")
|
| 359 |
fps_fname = save_hdf(morgan, molnames, molpath)
|
| 360 |
mol_imgs = draw_molecules(smiles)
|
| 361 |
+
mol_features, mol_ids = main(mol_index, model_path, model_type, mol_path=molpath, image_resolution=image_resolution)
|
| 362 |
predefined_features = False
|
| 363 |
else:
|
| 364 |
mol_index = pd.read_csv(mol_index_file)
|
|
|
|
| 375 |
print(img_features.shape)
|
| 376 |
print(mol_features.shape)
|
| 377 |
|
| 378 |
+
top_n = int(top_n)
|
| 379 |
+
|
| 380 |
logits = img_features @ mol_features.T
|
| 381 |
mol_probs = (30.0 * logits).softmax(dim=-1)
|
| 382 |
+
top_probs, top_labels = mol_probs.cpu().topk(top_n, dim=-1)
|
| 383 |
|
| 384 |
# Delete this if want to allow retrieval for multiple images
|
| 385 |
top_probs = torch.flatten(top_probs)
|