Attention Mechanism integrated with Unet Network was adopted for polyp segmentation task. The weights were transfered from a UNet network trained over brain MRI scans. The paper was submitted to Medico Challenge 2020. .

The largest connected component clearly possesd community structure of the co-authorship network. We also found that the communities inside the connected component had fairly good density and authors had strong bonding inside the detected communities.

As our dataset contained slices of the entire abdominal region and were not annotated at all, region growing approach based on morphological operation was chosen to segment the kidney.

In today’s internet world, comment sections should be carefully handled by companies. This section can be full of toxic and intimidating comments in the absence of a moderator. Human moderator along is not a practical solution for gate keeping the toxic comments.In this project, attempt was made to improve the classification of toxic online comments using the dataset from kaggle competition.

Not all citations are equally important, there is a need of deeper analysis of citation features to assess the impact of a research paper and identify a citation if it is important or non-important.

Miscellaneous - Fav Books  /  Travel Trace  /  Hobbies  /