Second Position at Innerve Data Hackathon by IGDTUW

Our team BitFlip's submission "Semen Analysis with Deep Learning for IVF and Fertility Labs" bagged second prize in the national level data hackathon Innerve organized by Indira Gandhi Delhi Technical University for Women 🥳🥳

Huge shout-out to my team members Priyansi Singh and Biswaroop Bhattacharjee.

Current methods of semen analysis are time consuming and prone to human error. We decided to build a deep learning based solution that could predict some parameter of semen quality from a video. We used the VISEM dataset and trained spatial-temporal 3D Conv Nets to predict the sperm motility from microscopic videos of the semen sample.

Finally we used 3D ResNets (18 and 34) to train 3 models that gave really good loss and accuracy. We improved upon the latest research (https://lnkd.in/dpvbfKt) by using 3D CNNs as they perform better on video data.

Our solution turned out to be dramatically cost and time effective due to it's digital nature. ✌️✌️

Code and Presentation : https://gitlab.com/innerve/sperm-motility-analysis