search

abhisheks008 / DL-Simplified

Deep Learning Simplified is an Open-source repository, containing beginner to advance level deep learning projects for the contributors, who are willing to start their journey in Deep Learning. Devfolio URL, https://devfolio.co/projects/deep-learning-simplified-f013
https://quine.sh/repo/abhisheks008-DL-Simplified-499023976
MIT License
345 stars 295 forks source link

Global Land Cover Mapping using Image Processing #813

Closed ArismitaM closed 3 months ago

ArismitaM commented 3 months ago

Pull Request for DL-Simplified 💡

Issue Title: Global Land Cover Mapping using Image Processing

Closes: #503 #issue number that will be closed through this PR

Describe the add-ons or changes you've made 📃

I have made a folder containing all the required files and subfolders. I have tried and tested 3 models for this issue, namely YOLOv5, RetinaNet, and VGG16, and added the best model.

Give a clear description of what have you added or modifications made

I have added a new folder to this repo with the same name as that of the issue name which is 'Global Land Cover using Image Processing'. This was my first time working with .tif images and I have used the DBScan clustering technique to draw bounding boxes and then trained each of the models.

Type of change ☑️

What sort of change have you made:

How Has This Been Tested? ⚙️

Describe how it has been tested There was a folder that contained images for testing, so, I ran testing on those images, and based on the results I had to re-train some of the models with varying epochs and batch sizes to get the best result.

Describe how have you verified the changes made I saw the testing on the images where the model was able to classify most of the data correctly. I trained it with 5000 epochs. I verified the testing with the help of the mAP and epoch loss graphs.

Checklist: ☑️

github-actions[bot] commented 3 months ago

Our team will soon review your PR. Thanks @ArismitaM :)

abhisheks008 commented 3 months ago
  1. Please upload the .ipynb file inside the Model folder. It's the optimum file for this project.
  2. Minor rename changes are required, readme should be README.md.
ArismitaM commented 3 months ago

I have uploaded the .ipynb file inside the model folder and have renamed the README.md file

abhisheks008 commented 3 months ago

I asked you to upload the whole code inside the notebook, how you implemented all the three models, how you made those graphs, how you implemented the architecture all those codes should be written inside the notebook file. You can check out the existing projects for your reference.

ArismitaM commented 3 months ago

I have changed my .ipynb file. I have added the training graphs and the testing.

abhisheks008 commented 3 months ago

Hi @ArismitaM sorry for the late reply. I have seen your model.ipynb file, I didn't find the other two model implementations apart from YoloV5. Can you explain this to me?

ArismitaM commented 3 months ago

Hi @ArismitaM sorry for the late reply. I have seen your model.ipynb file, I didn't find the other two model implementations apart from YoloV5. Can you explain this to me?

Hello @abhisheks008, I thought we were supposed to add only the best model initially, so I added only the YOLO model. I have added the RetinaNet model too. Coming to VGG16, I realized that it was ideal for detection and not for classification, therefore it cannot detect more than 1 class. But my training has 8 classes so VGG was not ideal. I have mentioned this in the README.md file too.