Prateek Keserwani is working as a Lead Engineer (Machine Learning) in a Samsung Research Institute Bangalore, India. He did Ph.D. in Dept. of CSE, IITR, advised by Dr. Partha Pratim Roy (IIT Roorkee).on the detection and concealment of text in natural images. He completed his M.Tech from the University of Allahabad, Allahabad (UP) in 2015. He has also served as a guest faculty in NIT Allahabad (from 2010 to 2012) and the University of Allahabad (2015 - one semester). He also acted as a reviewer in many journals, conferences, and Workshops.

Currently, he is working in the area of neural architecture search. The neural architecture search is a step toward Learn2Learn. The current deep learning method can learn automatic features but needs the handcrafted design of the architectures. But, the neural architecture search obtains the optimal neural network from the search space and eliminates the need to design the architectures manually.  

Reviewer

Journals

• IEEE Transactions on Circuits and Systems for Video Technology
• ACM Transactions on Asian and Low-Resource Language infomation Processing
• IEEE Access
• IET Image Processing
• IET Biometrics
• Multimedia Tools and Applications
• International Journal on Document Analysis and Recognition
• The Visual Computer
• SN Computer Science

Conferences

• ICPR 2022
• ICFHR 2020
• CVIP 2021
• Workshop ECCV 2022

Talk and Tutorials

• Talk on Convolutional Neural Network using PyTorch, a short term course on Deep Learning and its Application, organized by the EICT Academy in IIT Roorkee.
• Tutorial on Multi-layer Perceptron and Convolutional Neural Network using Tensorflow in a course of "Deep Learning, Image processing with Computer Vision" for BEL engineers in 2021. 
• Tutorial on Scene Object Detection in WCVIP 2020.

Fellowship

• VVS-Fellowship for Ph.D. from the Ministry of Electronics and Information Technology (MeitY)
• MHRD fellowship for M.Tech

Competitions

• Our team (PARIMAL) ranked third in the ICDAR Large Structured Chinese Family Records Layout (Task 2). [github] [Competition Paper] [breif report]
• Our team (PARIMAL) ranked fifth in the Semantic segmentation competition on IDD Lite Dataset organized by the IIIT Hydrabad.[github] [Competition Page]

Selected Publications

Journals 

• Keserwani, P., Saini, R., Liwicki, M., & Roy, P. P. (2022). Robust Scene Text Detection for Partially Annotated Training Data. IEEE Transactions on Circuits and Systems for Video Technology, 32(12), 8635-8645.
• Keserwani, P., & Roy, P. P. (2021). Text region conditional generative adversarial network for text concealment in the wild. IEEE Transactions on Circuits and Systems for Video Technology, 32(5), 3152-3163.
• Keserwani, P., Dhankhar, A., Saini, R., & Roy, P. P. (2021). Quadbox: Quadrilateral bounding box based scene text detection using vector regression. IEEE Access, 9, 36802-36818.

Conferences

• Lotey, T., Keserwani, P., Wasnik, G., & Roy, P. P. (2022, August). Cross-Session Motor Imagery EEG Classification using Self-Supervised Contrastive Learning. In 2022 26th International Conference on Pattern Recognition (ICPR) (pp. 975-981). IEEE.
• Keserwani, P., De, K., Roy, P. P., & Pal, U. (2019, September). Zero shot learning based script identification in the wild. In 2019 international conference on document analysis and recognition (ICDAR) (pp. 987-992). IEEE.
• Keserwani, P., Ali, T., & Roy, P. P. (2017, November). TRPN: A text region proposal network in the wild under the constraint of low memory GPU. In 2017 4th IAPR Asian Conference on Pattern Recognition (ACPR) (pp. 286-291). IEEE.