Patents
- METHOD AND SYSTEM FOR OPTIMIZING POWER CONSUMPTION IN A DISPLAY DEVICE
[ patent]
Ashish Kumar SINGH, Shubham PATERIA, Kumar KATRAGADDA, Krishna Kishor JHA, Vaisakh Punnekkattu Chiryal Sudheesh BABU. Samsung R&D India-Bangalore.
[IN Patent] India Patent 201741038214, 2019.
- METHOD AND SYSTEM FOR AN EYE SENSATION PREDICTION BASED DISPLAY ENHANCEMENT
[ patent]
Krishna Kishor JHA, Ashish Kumar SINGH, Deepthy RAVI, Shubam PATERIA, Vaisakh Punnekkattu Chirayil SUDHEESH BABU, Mahammadrafi Raimansab MANIYAR. Samsung R&D India-Bangalore.
[IN Patent] India Patent 201741008468, 2019.
Other Projects
- Power efficient, bandwidth optimized and fault tolerant sensor management for IOT in Smart Home
Pradulla Choubey, Shubham Pateria, Aseem Saxena, Vaisakh PCSB, Krishna Kishor Jha, Sharana Basaiah PM. Samsung R&D India-Bangalore.
[IEEE IACC] In 2015 IEEE International Advance Computing Conference (IACC), Banglore, India, 2015, pp. 366-370, doi: 10.1109/IADCC.2015.7154732.
- Aspect Based Sentiment Analysis using Sentiment Flow with Local and Non-local Neighbor Information
[ paper]
Shubham Pateria
[COLING] In Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers (pp. 2635-2646).
- Character-level Sign Language Interpretation from Hand-gestures using Recurrent Convolutional Neural Networks
[ Project]
Shubham Pateria, Abhay MS Aradhya, Andri Ashfahani, Mohamed Ragab. NTU, Singapore.
We developed a Convolutional LSTM model that takes a sequence of hand gestures as input and produces the corresponding word as output. The gesture data was obtained from Sign Language MNIST on Kaggle.
Our Conv LSTM model achieved a word prediction (test) accuracy of 96.8% compared to the 64.5% accuracy achieved only with character prediction CNN (without hidden state recurrence).
On top of this, we also developed a Sequence-to-Sequence LSTM which was trained on the pairs of (incorrectly spelled, correctly spelled) words. This was used as a post-processor on the output of Conv LSTM. This improved the test accuracy to 99.1%.
|