kinshukgoel4 / metadata-engine Goto Github PK

There are a few security vulnerabilities due to the versions of specific libraries like pillow, torch. Upgrade the library version in requirements.txt file to fix it

Improve Contribution Guidelines in Contribution.md

Example:

1. Add Cloning through HTTP git clone https://github.com/<YOUR-USERNAME>/Metadata-Engine.git
2. Add Cloning through SSH git clone [email protected]:<YOUR-USERNAME>/Metadata-Engine.git
3. Improve the Commit Message in this way git commit -m "feat/docs/fix: :emoji-name: <EXPLAIN-YOUR_CHANGES>"
4. While pushing mention issue number git push origin fix-issue-<ISSUE-NUMBER>

Since the video frames are processed one by one in a loop, this process could potentially be sped up by using multiple threads to process the frames in parallel.

Detect Object metadata such as :

• object_id
• object_class
• object_confidence
• bounding_box

Add different badges at starting of readme such as:

• Open Source
• Contributors
• Maintained

The code is not working properly and is not yielding output metadata files for multiple raw input video files

Instead of hardcoding the video paths and metadata directories, you could add command line arguments to allow the user to specify them when running the script. This would make the script more flexible and easier to use in different contexts.

Send the status of Metadata generation to the user's WhatsApp

Update/Style the Setup⚒ Instructions in README.md

Instead of saving the metadata as a JSON file, modify the Engine class to save the metadata in a database. This would allow for more efficient querying and analysis of the metadata, especially while processing a large number of videos.

Integrate speech_to_text.py in main.py file

To simplify deployment and setup create a Docker container

Make it more user-friendly readable by:

• Adding Emojis
• Adding GIF's
• Adding Batches

The requirements.txt and environment.yml aren't up to date with the latest code and libraries

You could add a filter to the Engine class to only include objects of a certain class in the metadata output. For example, if you only wanted to detect people, you could filter out all other object classes.

Append the Contribution section of the readme to the Contribution.md file

To get started Please visit: https://trykino.notion.site/Kino-AI-Challenge-Metadata-Engine-Hackathon-bbf5b2ec6c904a86a4ff3ecddd57e320#45bbc7bf17e144b7bd08569c24c4806b

Detect metadata fields such as :

• frame width in pixels
• frame height in pixels
• the frames per second(fps) of the video
• the codec used to encode the video

You could modify the Engine class to also draw bounding boxes around the detected objects in each frame and save the resulting frames as images or a new video. This would provide a way to visually verify the accuracy of the object detection.

Add sections such as:

• Contribution Guidelines
• Code of Conduct
• LICENSE

The Readme.md file is not upto the date and requires updates as per the new code/commits made to the main.py file.

EX: Recently object detection was implemented and integrated with person detection and pushed into main branch of the main.py file but https://github.com/kinshukgoel4/Metadata-Engine#metadata-generated is not updated with new metadata fields

Make a initial commit with README and LICENSE

If you were processing a large number of videos, to parallelize the video processing by using a library like joblib or multiprocessing. This would allow to process multiple videos simultaneously, speeding up the overall processing time.

Remove Contribution guidelines from readme and add a hyperlink to Contribution.md

Write bullet points rather than long para in README.md file