https://github.com/bryanoliveira/bone-age-regression
A Deep Learning approach to solve the bone age regression problem.
https://github.com/bryanoliveira/bone-age-regression
pytorch torchvision
Last synced: 27 days ago
JSON representation
A Deep Learning approach to solve the bone age regression problem.
- Host: GitHub
- URL: https://github.com/bryanoliveira/bone-age-regression
- Owner: bryanoliveira
- Created: 2020-01-16T23:04:28.000Z (over 5 years ago)
- Default Branch: master
- Last Pushed: 2021-05-23T00:39:29.000Z (over 4 years ago)
- Last Synced: 2025-02-22T09:48:08.511Z (7 months ago)
- Topics: pytorch, torchvision
- Language: Python
- Homepage:
- Size: 5.29 MB
- Stars: 3
- Watchers: 3
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Bone Age Regression with Deep Learning
This is my code for the [I2A2 Bone Age Regression competition](https://www.kaggle.com/c/bone-age-regression). I learned a lot by building this pipeline from scratch and experimenting with different model architectures and optimizers. This was my first end-to-end image regression model, and it was very nice seeing my theoretical knowledge work in practice.
This competition was inspired by [RSNA's Bone Age challenge](https://www.kaggle.com/kmader/rsna-bone-age), in which given hand X-ray images, the model should predict the patient's bone age.
> X-ray images provided in the competition's dataset.
My final solution used a [ResNet50](https://arxiv.org/abs/1512.03385) architecture, a [Rectified Adam](https://arxiv.org/abs/1908.03265) optimizer and geometric data augmentations. This model achieved a Mean Average Error of 13.2 after 20 epochs of training, which I believe could be improved given more training time and a better preprocessing pipeline (e.g. using object detection to segment the hands and normalizing hand rotation). Unfortunately, I didn't save all the hyperparameters I experimented with (neither their results), but you'll find the ones I used for my last submission in the code.
I used [tensorboard](https://www.tensorflow.org/tensorboard) to log the training curves and [tqdm](https://github.com/tqdm/tqdm) to track progress. I also used [FCMNotifier](https://github.com/bryanlincoln/fcm-notifier), a tool I made to send logs as notifications to my phone.
## Requirements
See [requirements.txt](https://github.com/bryanlincoln/bone-age-regression/blob/master/requirements.txt).
## Usage
- Download the requirements with `pip install -r requirements.txt`
- Download the dataset and sample submission with `sh download_data.sh`. You may need to log in with your Kaggle account in order to do it.
- Train the ResNet50 model with `python boneage.py`
- Try different models and hyperparameters by editing the training script or use the `boneage.ipynb` notebook to do it interactively.
## Credits
I used the vision models already [implemented in torchvision](https://github.com/pytorch/vision/tree/master/torchvision/models) with slight changes. You can try other torchvision models by adding the `in_channels` parameter to generalize the number of input channels since torchvision models work with RGB images.