spaceml-org / image-similarity-search Goto Github PK

This is a built in library and can be removed from requirements

As title

Can you share code for training a model (like samples/uc_merced.pt), please?

• Get rid of the .DS_Store file.
• Update .gitignore to follow the python template.
• Update requirements.txt using the following command within your virtual env:
  conda env export requirements.txt
  Would be better to create a new virtual environment for this so that unnecessary packages are not included in the requirements.txt files.
• Move source files to the src directory. When this is converted to a python package, it will contain just the setup.py file in the main directory. (Can be done at the time of creating the python package as well)

--- EDIT ---

• for .gitignore use this link instead.

• If it is not critical and does not break things, move app.py code to the main execution.

if __name__ == "__main__":
    <<code>>

• Get rid of the imports that are not being used (both files).
• Run a code-formatter to have the code follow pep8 standards (both files). I use Black.

Section 1: First statement/paragraph

• Have a title such as "What is Image Similarity Search" or "Introduction" for the very first statement/paragraph.
• Explain the first paragraph better. The current statement is not very elaborate or clear. Make sure to clearly specify the Input, the Output, and the aim/action of the pkg.

Section 2: How it works?

• End the title with a question mark.
• Involve our design expert for the image/diagram.
• Point 3, highlight "n" in the "n closest images". Can use tilde or capital case.
• Have numbered points instead of bullets. These numbers will correspond to the different labeled sections of the image, giving a visual idea of what different points are.

Section 3: Usage

• Have a hyperlink to streamlit website on the word "streamlit".
• Have a subsection within the first statement that says, "If you do not already have streamlit installed, follow these steps first: ..." followed by the installation steps.
• Have screenshot of the streamlit app running.

Section 3.1: Steps

• Again, correspond the numbers in the Steps section to the marked numbers on the streamlit screenshot mentioned in the previous point. Can decide whether single screenshot serves better for multiple points or a single screenshot for each point.
• Involve our design expert for ideas/suggestions.

Section 3.2: Samples

• Have a hyperlink to the samples folder with a smaller readme inside that directory. The smaller readme can have the same information as this subsection in the main README.
• Mention the "download dataset" instruction as a different statement. Also, mention the preferred location to download the dataset.
• "Default embedding size is 21." specify this as 21 float values or 21x1 float values.

Section 4: Dependencies

• Move this section either above the Usage section or as a subsection inside the Usage section.
• Mention the pip install command in a separate line similar to the wget or streamlit command mentioned in the previous sections.