[TODO]

[alessandro-maccario]

Check:

• https://www.quora.com/How-can-I-improve-my-validation-accuracy-in-my-CNN-model
• https://blog.keras.io/building-powerful-image-classification-models-using-very-little-data.html
• https://towardsdatascience.com/integration-of-dimension-reduction-methods-and-neural-network-for-image-classification-96281963fe24
• https://towardsdatascience.com/using-pca-to-reduce-number-of-parameters-in-a-neural-network-by-30x-times-fcc737159282
• https://wandb.ai/wandb_fc/tips/reports/How-to-Use-a-Learning-Rate-Scheduler-in-Keras--VmlldzoyMjU2MTI3
• Tuning the Hyperparameters and Layers of Neural Network Deep Learning: https://www.analyticsvidhya.com/blog/2021/05/tuning-the-hyperparameters-and-layers-of-neural-network-deep-learning/
• GRIDSEARCHCV: https://machinelearningmastery.com/grid-search-hyperparameters-deep-learning-models-python-keras/
• https://studymachinelearning.com/keras-imagedatagenerator-with-flow_from_directory/

References:

• https://medium.com/free-code-camp/everything-you-need-to-know-to-master-convolutional-neural-networks-ef98ca3c7655
• https://medium.com/mlearning-ai/computer-vision-exploring-your-image-datasets-the-right-way-538c5ae8ca5d

CNN EXPLAINED

• https://www.youtube.com/watch?v=pj9-rr1wDhM

CNN

• https://machinelearningmastery.com/how-to-configure-image-data-augmentation-when-training-deep-learning-neural-networks/

DATA AUGMENTATION

• https://www.analyticsvidhya.com/blog/2020/08/image-augmentation-on-the-fly-using-keras-imagedatagenerator/#:~:text=Keras%20ImageDataGenerator%20is%20a%20gem,up%20on%20the%20overhead%20memory!
• https://www.analyticsvidhya.com/blog/2021/06/offline-data-augmentation-for-multiple-images/
• https://github.com/AISangam/Image-Augmentation-Using-OpenCV-and-Python/blob/master/Image%20Augmentaion%20Part1.py
• https://towardsdatascience.com/complete-image-augmentation-in-opencv-31a6b02694f5

HISTOGRAMS FOR COLORS IN IMAGE

• https://pyimagesearch.com/2021/04/28/opencv-image-histograms-cv2-calchist/
• https://web.stanford.edu/class/cs101/image-1-introduction.html#:~:text=Each%20of%20the%20red%2C%20green,in%20a%20shade%20of%20orange.

MANIFOLD

• https://towardsdatascience.com/manifolds-in-data-science-a-brief-overview-2e9dde9437e5

DROPOUT

• https://stackoverflow.com/questions/47892505/dropout-rate-guidance-for-hidden-layers-in-a-convolution-neural-network

FINE-TUNING

• https://towardsdatascience.com/fine-tuning-a-cnn-model-for-image-classification-a886e4b539b3
• https://www.analyticsvidhya.com/blog/2021/05/tuning-the-hyperparameters-and-layers-of-neural-network-deep-learning/
• https://machinelearningmastery.com/grid-search-hyperparameters-deep-learning-models-python-keras/

CROSS-VALIDATION KERAS

• https://towardsdatascience.com/increase-the-accuracy-of-your-cnn-by-following-these-5-tips-i-learned-from-the-kaggle-community-27227ad39554

GRIDSEARCH

• https://www.youtube.com/watch?v=TvB_3jVIHhg

TO TRY:

1. Get all the lists of loss and accuracy
2. get the mean for each values
3. plot them!!!

TO CHECK FOR MORE IDEAS FOR THE ARCHITECTURE

• https://machinelearningmastery.com/review-of-architectural-innovations-for-convolutional-neural-networks-for-image-classification/
• https://levelup.gitconnected.com/simple-image-classification-with-cnn-dd5ee3b725
• https://www.learndatasci.com/tutorials/convolutional-neural-networks-image-classification/
• https://www.guru99.com/convnet-tensorflow-image-classification.html

References:

• https://github.com/TetsumichiUmada/food_app

  • https://github.com/TetsumichiUmada/food_app/blob/master/1_food_app.ipynb
  • https://github.com/Pyligent/food101-image-classification/blob/master/food-101_EDA_Model_RestNet_final.ipynb
  • https://github.com/TetsumichiUmada/food_app/blob/master/0_dataset.ipynb

• https://github.com/AhmedMaghawry/Food-Image-Recognition/blob/master/Food_Detection.ipynb

• FoodApp: http://blog.stratospark.com/creating-a-deep-learning-ios-app-with-keras-and-tensorflow.html

• https://medium.com/@ksusorokina/image-classification-with-convolutional-neural-networks-496815db12a8

• http://blog.stratospark.com/creating-a-deep-learning-ios-app-with-keras-and-tensorflow.html

• Rejection class:

  • https://hal.archives-ouvertes.fr/hal-02963115/document
  • https://ai.stackexchange.com/questions/24914/image-classification-need-method-to-classify-unknown-objects-as-trash-3d

• CNN:

  • https://medium.com/free-code-camp/everything-you-need-to-know-to-master-convolutional-neural-networks-ef98ca3c7655

-None of the above class approach:

• https://arxiv.org/pdf/1511.03719.pdf

• Cat/Dog dataset:

  • https://www.kaggle.com/competitions/dogs-vs-cats/data?select=train.zip
  • https://data-flair.training/blogs/cats-dogs-classification-deep-learning-project-beginners/

• Data augmentation:

  • https://medium.com/mlearning-ai/albumentations-a-python-library-for-advanced-image-augmentation-strategies-752bff3a3da0

Check the following notebook done during class:

• _FDLE04.ipynb

Optimizers:

• https://towardsai.net/p/l/impact-of-optimizers-in-image-classifiers
• https://opt-ml.org/papers/2021/paper53.pdf
• https://keras.io/api/optimizers/
• https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8321140/
• https://www.lightly.ai/post/which-optimizer-should-i-use-for-my-machine-learning-project
• https://medium.com/geekculture/a-2021-guide-to-improving-cnns-optimizers-adam-vs-sgd-495848ac6008

Refactoring the code:

• USE TYPE ANNOTATION:

  • https://towardsdatascience.com/type-annotations-in-python-d90990b172dc
  • https://blog.logrocket.com/understanding-type-annotation-python/

• CONVERT REUSABLE CODE IN FUNCTIONS AND MOVE THEM TO THE UTILS.PY FILE

• YOU SHOULD ALSO SAVE THE FINAL DATASET WITH ALL THE PERFOMANCE MEASURES AND TIME SPENT FOR EACH MODEL:

  • https://github.com/alessandro-maccario/Industry-Lab-project-private/blob/main/Notebooks/Part2_Training.ipynb

[alessandro-maccario]

• [ ] Addizionare PCA nel transfer learning, unicamente come metodo di comparazione in termini di riduzione dei parametri. Non sarà il modello che poi utilizzeremo, ma solo come comparazione!

• [ ] Salva tutti i modelli e passa a migliorare il notebook di evaluation.

• [ ] Dopodiché riprendi l'ultimo notebook di Industry Lab come riferimento per il notebook di testing e deployment!

• [ ] Inserisci valore di cut off sulla percentuale se inferiore al 50% allora la risposta fornita dalla rete deve essere relativa alla classe di rigetto.

• [ ] Per il deployment usa i seguenti riferimenti:

  https://www.freecodecamp.org/news/how-to-deploy-your-machine-learning-model-as-a-web-app-using-gradio/ https://www.section.io/engineering-education/deploying-machine-learning-model-as-an-app-in-python-using-gradio/ https://python.plainenglish.io/build-and-deploy-machine-learning-apps-with-gradio-the-easiest-way-to-create-a-web-user-interface-6f5219025da3 https://medium.com/@kalyaniavhale7/tutorial-on-gradio-library-ecb8055923a1 https://abidlabs.medium.com/deploy-a-machine-learning-model-in-seconds-56c400ecc0b3 https://trojrobert.github.io/a-guide-for-deploying-and-serving-machine-learning-with-model-streamlit-vs-gradio/ https://abdulsamodazeez.com/how-to-build-a-machine-learning-web-app-in-python-using-gradio https://www.youtube.com/watch?v=53x330upumY