KKKaustubh Rajendra Kulkarni

Kaustubh Rajendra Kulkarni

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Kaustubh Rajendra Kulkarni has a total of 17 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Social Media Analytics for Monsoon Trends in India: A Multi- Platform NLP Approach

    by Kaustubh Rajendra Kulkarni, Pradnya Patil, Priyanka Sharma

  • Open Access

    Transfer Learning for Agricultural Disease Recognition: Performance Evaluation of Modern CNN Architectures

    by Saundarya Subramaniam, Shalini Majumdar, Shantanu Nadar, Kaustubh Rajendra Kulkarni

    Agricultural productivity faces significant challenges from plant diseases, particularly in resource-constrained environments where traditional diagnostic methods are inadequate. This study evaluates the effectiveness of modern convolutional neural network (CNN) architectures for automated plant disease identification through transfer learning methodologies. We systematically compare five distinct approaches: a custom-designed CNN alongside four pre-trained models (EfficientNetB0, ResNet101, InceptionV3, and MobileNetV2) using a comprehensive dataset of over 130,000 plant images. Interestingly, EfficientNetB0 achieves superior performance with 97.10% validation accuracy, while our custom CNN architecture attains competitive results at 95.76% accuracy. The findings indicate that transfer learning approaches significantly enhance model performance while reducing computational requirements, making them particularly suitable for deployment in agricultural settings with limited technological infrastructure. This research contributes to the growing body of knowledge on AI-driven agricultural solutions and provides practical insights for implementing scalable disease detection systems.

  • Algorithmic Issues, Challenges, and Theoretical Concerns of ChatGPT

    by Pradnya Patil, Kaustubh Kulkarni, Priyanka Sharma

    Large language model (LLM) ChatGPT has made tremendous progress in natural language processing (NLP), especially in human-quality text production, multilingual translation, and content creation. However, because of its extensive use, algorithmic issues, challenges, and theoretical queries come up. Fairness, bias, explainability, generalization, originality, inventiveness, and safety are the main topics of this study's examination of ChatGPT's intricate theoretical and algorithmic components. It looks into the possibility of explainability, transferability, generalization, bias in the data, and the model's capacity to provide original and imaginative content. It also covers possible issues including harmful use, disseminating incorrect information, and offensive or misleading content. These limitations can be addressed so that ChatGPT can be improved to provide LLMs that are more dependable, accountable, and long-lasting while posing no needless risks.


  • Open Access

    Changes in Facial Expression as Biometric: A Database and Benchmarks of Identification

    by Rain Eric Haamer, Kaustubh Kulkarni, Nasrin Imanpour, Mohammad A. Haque, Egils Avots, Michelle Breisch, Kamal Nasrollahi, Sergio Escalera, Cagri Ozcinar, Xavier Baro, et al.

    Facial dynamics can be considered as unique signatures for discrimination between people. These have started to become important topic since many devices have the possibility of unlocking using face recognition or verification. In this work, we evaluate the efficacy of the transition frames of video in emotion as compared to the peak emotion frames for identification. For experiments with transition frames we extract features from each frame of the video from a fine-tuned VGG-Face Convolutional Neural Network (CNN) and geometric features from facial landmark points. To model the temporal context of the transition frames we train a Long-Short Term Memory (LSTM) on the geometric and the CNN features. Furthermore, we employ two fusion strategies: first, an early fusion, in which the geometric and the CNN features are stacked and fed to the LSTM. Second, a late fusion, in which the prediction of the LSTMs, trained independently on the two features, are stacked and used with a Support Vector Machine (SVM). Experimental results show that the late fusion strategy gives the best results and the transition frames give better identification results as compared to the peak emotion frames.