Fake Visuals Recognition: Human vs Computer
This objective of this task is to assess the capacity of humans and machines in perceiving counterfeit visuals.
1. Built up a synthetic image generator utilizing a Generative Adversarial Network that was trained on a large number of real human pictures.
2. Prepared a Convolutional Neural Network(CNN) classifier that recognizes genuine and counterfeit visuals
3. Directed a review from volunteers to recognize genuine and counterfeit visuals.
The outcomes inferred that the CNN classifier was better at arranging counterfeit visuals with a precision of 91% over people at 41%. The task won the third spot for "Famous Vote for Best Project Award" in the Crowdsourcing in Computer Vision class.
​
Database: SQL
Algorithms: Nvidea Style GAN Architecture, Inception V3 (for CNN).
Contact Volume Prediction Model.
The objective of this task is to build an Intelligent ML Contact Volume forecasting model for optimum headcount planning at Dell.
1. Improved Models: Develop enhanced ML models to predict call volume.
2. Scalability: Design ML models that can dynamically predict call volume with the vision to minimize manual intervention.
3. Data Integrity Checks: Evaluation of models on multiple standard metrics to determine the optimal solution.
4. FbProphet was used to Improve prediction for >75% queues with an average of 30% error reduction.
​
Database: SQL
Algorithms: FbProphet, Deep Neural Networks.