I have co-founded Hyperint. We have 2 products live. Selfie Validator and WordOfMouth. Working on one more product.

Ashish Gupta  • Peerlist

In this image is captured using  open CV. Neural Net is simple CNN architecture in Pytorch and dataset used is Kaggle Flower dataset with 11 classes of flowers. For web Searching Web Browser and GoogleSearch API is used. Training is done in Google Colab GPU.


Botanical Flower Detector with Web Search Facility

It provides one to one with group Chat option.GUI is designed with swings and Multithreading is the most crucial part of this project.

Chat Application Using Socket Programming in JAVA

Filtering of the dataset is the crucial Part of this project. Single Month Dataset(July) is used and interpreted using SQLite python.  Some Rules(logic) are used for filtering of the dataset.
I used Neural Machine Translation Concept. Next, it also contains support for attention mechanisms,

Chatbox using RNN(Bidirectional RNN's) with attention machan

Drug Discovery using GNNs

For 3D image reconstruction,  extracting silhouette using deep Residual network in pytorch. Dataset used was PascalVoc 2012.

Extracting Silhouette By Semantic Segmentaion

In this project what I basically follow are LSTM's and GRU's(RNN). The data was created by the European Union which translates a lot of their communications to the languages of the member countries of the European Union. The project is performed in Tensorflow with Keras.

Neural Machine translation NLP

Using ShapeNet and Pascal Voc dataset we estimation rotation and translation(camera extrinsic parameters).
Using CNN as pose regressor and the project is implemented in pytorch.I used L1 loss and Adam optimizer with a learning rate of 1e-4.

Object Pose Estimation(6 DOF)

Here in this, an online portal is created using Java Server Pages with HTML and CSS.Database-MySql
Features:
1. Login/accounts for students and Faculty.
2. Faculty can conduct various examinations(MCQ's) for particular batches.
3. Students marks are evaluated and updated in the DataBase.
4. At the end of the sem student marks are evaluated using Fuzzy Logic and marks sheet is generated.

Online Test System with Marks Evalution Using Fuzzy Logic(On

It's a GUI Java-based game which uses various Heuristic(Artificial Intelligence). The set of rules are designed on two bases.
1. Winning Condition-The winning goes in the hand of the computer player, ie. now whatever move human will going to take, there is no chance of winning of the human player.
2.Not in winning condition-The computer player is not in the winning condition, so it will try to take all those moves to let the computer player be in winning strategy.

Rule based NIM Game

Here in this, I used the Recurrent Neural Network with 3 recurrent units. Used GRU's and used a data-set consisting of 50000 reviews of movies from IMDB.
The code is in Tensorflow with Keras.
accuracy training: 0.8765 
validation accuracy: 0.8752


Sentiment analysis using RNN(GRU's) (NLP)

I am using deep Residual Nets(CNN) and the datasets using are PascalVoc 2012(Real Images) and Shape Net(Synthetic Images).
Papers I am following:
1.Deep Single-View 3D Object Reconstruction with Visual Hull Embedding 
2. 3D-R2N2: A Unified Approach for Single and Multi-view 3D Object Reconstruction,