About me |
|
Recent Updates |
Pushkar Gole is a passionate researcher and academician having Ph.D. degree from the Department of Computer Science, University of Delhi. With an
M.Sc. in Computer Science from the same institution and a B.Sc. (H) in Computer Science from Keshav Mahavidyalaya,
University of Delhi, He is dedicated to exploring cutting-edge research areas. His interests span AI in Agriculture,
Computer Vision, Digital Image Processing, Blockchain, and Deep Learning.
For more details, please refer my CV. |
|
 June 11, 2025:
April 30, 2025:
|
Developed an Android mobile application named PlantD2R2S-Lite which can detect plant diseases by capturing the leaf images through camera of mobile phone. This application can also segment the diseased areas of infected leaf image and estimate the severity of identified disease by calculating percentage of diseased pixels out of the sum of healthy and diseased pixels. After identifying plant disease and estimating its severity, the mobile application generate disease remdiation advisory. The developed application can work in both English and Hindi languages without Internet connectivity.
Developed a lightweight state-of-the-art framework named PDSE-Lite for plant disease identification and severity estimation by utilizng only two training samples per class of the dataset. This framework has been designed and developed by using Convolutional Auto Encoder (CAE) and Few-Shot Learning (FSL). As, this framework requires only two training instaces from each class, thus it alleviates the reliance on large annotated datasets which requires lots of human efforts.
Designed a novel lightweight Vision Transformer (ViT) model by replacing the Multi-Layer-Perceptron (MLP) module with the Inception module in the encoder of ViT model. The modified encoder block is named as Trans-Inception block. Replacing MLP module with Inception module reduces the number of trainable weight parameters and enhances the feature extraction abilities of ViT model. In proposed TrIncNet model, skip connections are also added around each Trans-Inception block which were not existed in the original ViT model. These skip connections prevents the model from overfitting. Moreover, eXplainable Artificial Intelligence (XAI) technique named LIME for generating human interpretable visual explanations for the predictions of proposed model.
Developed a lightweight hyrbid model based on CAE and CNN for diagnosing single plant disease from their digital leaf images. This model first obtains the compressed domain representations of leaf images using the encoder network of CAE and then uses these compressed domain representations for classification via CNN. Due to the reduction of spatial dimensions using CAE, the number of features and, hence, the number of trainable weight parameters of the lightweight hybrid model reduced significantly as compared to other state-of-the-art models.
Mentored the students of M.Sc. Computer Science in developing a smart contract for verifying the skills of applicants in the recruitment process. The major contribution of this system is that it allows the registered companies to monitor their employees and their skills. Furthermore, the proposed system will benefit the employees registering into the system to get their skills endorsed and experiences approved. Due to the need for achieving consensus in Blockchain technology, it is harder to make fraudulent profiles. By leveraging the security and decentralization offered by blockchain technology, the companies can evaluate the skills as well as experiences of the candidates before hiring them and be sure about the information verified by the candidate's former employers.
Developed a smart contract based framework in my M.Sc. major project for the management of Central Sector Scheme of Scholarship. This frameworks alleviates the different problems in the existing workflow of this scholarship like lack of traceability of application form, loss of application form in transit through Indian Postal Service, lack of transparency between students and their respective Education Boards, and lack of bank account verification.
Supervisor: Sr. Prof. Punam Bedi
Research Area: Use of Artificial Intelligence in Agriculture
Title of thesis: Lightweight and Few-Shot Image-based Plant Disease Diagnosis and Remedy Recommender System
Graduated with 79.95% Project: Smart contract for Central Sector Scholarship (CSS) Scheme
Graduated with 86.3%
Deep Learning, Computer Vision, Explainable AI
Java (Proficient), C (Prior experience), C++ (Prior experience), Python (Proficient), JSP (Prior experience), JavaScript (Prior experience), HTML (Prior experience), CSS (Prior experience), Solidity (Prior experience)
MySQL (Prior experience), Oracle (Prior experience), MongoDB (Prior experience)
VS Code, IntelliJ, PyCharm, Jupyter Notebook, Remix Studio, Android Studio, Ganache, Truffle
