search

OmdenaAI / myanmar-weather-forecasting

This project utilizes machine learning to predict extreme weather events and analyze traffic congestion in Southeast Asia. The repository includes data analysis, model training, and visualization scripts to forecast weather conditions for the next 30 days and enhance transportation efficiency.
GNU General Public License v3.0
11 stars 1 forks source link

Find sources for traffic congestion #2 #13

Closed CarlosGDCJ closed 1 year ago

CarlosGDCJ commented 1 year ago

We'll treat this problem as an object detection task. Meaning that we aim to develop a model that's capable of identifying cars/motorcycles in a picture.

In order to do that we need annotated (with bounding boxes) urban traffic data.

Finding something in Kaggle would be really valuable, as they provide access to GPUs, which are invaluable in the context of object detection.

MatthewDada commented 1 year ago

https://www.kaggle.com/datasets/sshikamaru/car-object-detection

https://www.kaggle.com/code/advaypatil/car-object-detection

Do these work?

CarlosGDCJ commented 1 year ago

We could use that. Can you look into this dataset @MatthewDada? Like checking some sample notebooks from Kaggle, seeing how the bounding boxes information is stored, how many classes are there, if there is any problem with the data, etc. and then giving a short summary.

MatthewDada commented 1 year ago

Tensorflow ain't exactly an area I know, but going through several notebooks there, I think someone with some knowledge about this field should be able to navigate, as the notebooks were well explained. some going all the way to train and test models with about 95% accuracy.

CarlosGDCJ commented 1 year ago

That's ok. Then I'll ask you to skim through the notebooks and select the ones (like 3-5) that look the most useful to us.

MatthewDada commented 1 year ago

Alright. Expect them tomorrow.

MatthewDada commented 1 year ago
  1. https://www.kaggle.com/code/shabbirjodhpurwala/major-project-5a7
  2. https://www.kaggle.com/code/prathameshchaudhary/car-object-detection-yolov5
  3. https://www.kaggle.com/code/mrsohelranapro/car-object-detection-yolov-x-models
  4. https://www.kaggle.com/code/eesuck/yolo-cars-detection

https://www.kaggle.com/code/advaypatil/car-object-detection/notebook. A different notebook (sent it with the first dataset link tho), but could help.

CarlosGDCJ commented 1 year ago

Nice one. I'll take a look at the notebooks.

CarlosGDCJ commented 1 year ago

@MatthewDada, I created #17 with a script to download the dataset from Kaggle and added you as a reviewer. I want to see if you can run the code without problems. Feel free to ask for changes/clarification on the PR page.