Review of master branch #7
Conversation
| @@ -0,0 +1,31 @@ | |||
| # Created by https://www.toptal.com/developers/gitignore/api/macos | |||
There was a problem hiding this comment.
You might want to add in Windows, Linux, and Python to this list.
| image = st.file_uploader( | ||
| "Upload Query Image", type=["png", "jpeg", "jpg", "tif"] | ||
| ) | ||
| submitted_1 = st.form_submit_button("Search") |
There was a problem hiding this comment.
What do you think of using submit_result as the variable name here?
| uploaded_image = create_temp_file(image) | ||
| image = read_image(image) | ||
| st.image(image) | ||
| # placeholder.empty() |
There was a problem hiding this comment.
Remove commented code.
| image = read_image(image) | ||
| st.image(image) | ||
| # placeholder.empty() | ||
| fig = st.session_state["indexer"].process_image(image, n_neighbors) |
There was a problem hiding this comment.
Would it explain it better if the variable name was something like processed_image instead?
|
|
||
|
|
||
| class Indexer: | ||
| # @st.cache(suppress_st_warning=True) |
There was a problem hiding this comment.
Remove commented line
| with torch.no_grad(): | ||
| data_matrix = ( | ||
| torch.empty(size=(0, size)).cuda() | ||
| if device == "cuda" |
There was a problem hiding this comment.
I don't quite understand this construct in Python. Is this an empty statement with an else clause that gets executed? If so, I'll recommend using something like:
if device is not "cuda":
torch.empty(size=(0, size))
| d = embeddings.shape[-1] | ||
| index = faiss.IndexFlatL2(d) | ||
| index.add(embeddings) | ||
| faiss.write_index(index, "index.bin") |
There was a problem hiding this comment.
Would it be useful to parameterise the index.bin file?
|
|
||
|
|
||
| def get_fig(neighbours, images_list): | ||
| fig, axarr = plt.subplots(1, len(neighbours), figsize=(7, 7)) |
There was a problem hiding this comment.
fig and axarr don't convey very much.
|
|
||
|
|
||
| def get_embedding(model, im): | ||
| def to_tensor(pil): |
There was a problem hiding this comment.
Is pil a PIL image? If so might be worthwhile naming it pil_image.
| def to_tensor(pil): | ||
| return torch.tensor(np.array(pil)).permute(2, 0, 1).float() | ||
|
|
||
| t = transforms.Compose( |
There was a problem hiding this comment.
I don't know what t here is. Might be worthwhile giving it a more descriptive name.
No description provided.