diff --git a/NN_last _report.pdf b/NN_last _report.pdf new file mode 100644 index 0000000..495ecbd Binary files /dev/null and b/NN_last _report.pdf differ diff --git a/README.md b/README.md index 45c501f..9603096 100644 --- a/README.md +++ b/README.md @@ -1,34 +1,28 @@ # Neural_Networks_Project The name of the project is the following: Augmenting Convolutional Networks with Attention-based Aggregation for Breast Cancer Detection. - -This is the final project for the course of Neural Networks 2021/2022 at Sapienza University of Rome. - +This is the final project for the course of Neural Networks 2021/2022 held by professors Aurelio Uncini and Danilo Comminiello, at Sapienza University of Rome. >Student: Filippo Betello Mat: `1835108`; > >Student: Federico Carmignani Mat: `1845479`; ## 📝 Assignment - -1. Reimplement the network architecture in the Paper [Link 🔗](https://arxiv.org/abs/2112.13692) for Image Classification on CIFAR10. +1. Reimplement the network architecture in the [Paper](https://arxiv.org/abs/2112.13692) for Image Classification on CIFAR10. 2. Apply this innovative architecture to Breast Cancer Detection. ## 💾 Dataset +- [CIFAR10](https://www.cs.toronto.edu/~kriz/cifar.html) +- Kaggle Dataset for Breast Histopathology Images [Link 🔗](https://www.kaggle.com/datasets/paultimothymooney/breast-histopathology-images) -- CIFAR10 [Link 🔗](https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz) - -## 📜 Report - -- power point presentation () -- paper [Link 🔗]() +## 📜 Results +For CIFAR10 dataset results are very good:

+

+For the Kaggle dataset we obtained 87.95% of accuracy:

+

+This work can be found in the [PDF report](./Neural_network_project_BETELLO_CARMIGNANI.pdf) and in the [PPT] presentation(./PPT_NN.pptx).>br>
- In these files you can read more about the code and the result of the project. - -## 💯 Final score: - -Score: `30L` +We noticed that in this last image we couldn't know if the attention map created was highlighting the correct patch of the image or not, so we decided to use another [dataset](https://wiki.cancerimagingarchive.net/display/Public/CBIS-DDSM) where the ground trouth were provided: +

+This last step is currently under developing by professor Comminiello and one of his PhD students. [Here](./NN_last_report.pdf) you can find an abstract. ## 🙋 Info - -for any doubt or clarification contact us on: - -- send an email at: carmignani.1845479@studenti.uniroma1.it or betello.1835108@studenti.uniroma1.it +For any doubt or clarification send an email at: carmignani.1845479@studenti.uniroma1.it or betello.1835108@studenti.uniroma1.it. diff --git a/Recap NN.xlsx b/Recap NN.xlsx deleted file mode 100644 index 853f704..0000000 Binary files a/Recap NN.xlsx and /dev/null differ diff --git a/src/Mass-Training_P_00001_LEFT_CC (1).png b/src/Mass-Training_P_00001_LEFT_CC (1).png new file mode 100644 index 0000000..e7bc748 Binary files /dev/null and b/src/Mass-Training_P_00001_LEFT_CC (1).png differ diff --git a/src/breast_2.png b/src/breast_2.png new file mode 100644 index 0000000..7a64ccf Binary files /dev/null and b/src/breast_2.png differ diff --git a/src/mio duck.png b/src/mio duck.png new file mode 100644 index 0000000..ce74646 Binary files /dev/null and b/src/mio duck.png differ