The projects will be divided into 3 main tracks:
For each task to solve we need create a issue, branch and pull request for validation of all team, also in each task to be solved by a person it is necessary to put the name of the user and the number of issue and pull request created at the end of the name of the task to solve in this description.
Data analysis: the main works in this tracks are:
[x] Find the distribution of data (Histogram, Use N-dims graphs like PCA or T-SNE)
[x] Find the relation between classes (association analysis)
[ ] Find the covariance and other metrics
Data Preprocessing
[x] Study the case of doing some preprocessing in the data
[ ] Find another dataset for pre train data (@lecasax, issue: #5 )
[ ] Use data augmentation (naive or more smart)
[ ] Normalize dataset
[ ] Create sub set of dataset for test models (dataset reduction, ideas: curriculum learning)
Train pipeline
[ ] Train with models pretrained from ImageNet (@so77id )
[ ] Create Predict code (@so77id, Issue: #4 )
[ ] Reduction of epoch time (@so77id )
[ ] Try to reduce the size of images (@so77id )
[ ] Use dropout for some layers (@so77id )
[ ] Test with different ensemble methods (voting, train FC, bosting, etc)
[ ] Try train the threshold parameter for prediction
The projects will be divided into 3 main tracks: For each task to solve we need create a issue, branch and pull request for validation of all team, also in each task to be solved by a person it is necessary to put the name of the user and the number of issue and pull request created at the end of the name of the task to solve in this description.
Data analysis: the main works in this tracks are:
Data Preprocessing
Train pipeline