A PhD scholar at the Institute of Mathematics and Computer Science, University of Sindh Syed Sohail Ahmed Shah presented his final seminar on research focusing on optimizing deep learning classifiers for tomato plant disease detection, including analysis of accuracy and training time trade-offs.

Vice-Chancellor Prof Dr Fateh Muhammad Marri presided over the seminar. The scholar highlighted key aspects of improving model performance and efficiency in agricultural disease detection.

He shared key findings of his research and noted that deep learning models, particularly convolutional neural networks could accurately detect and classify tomato plant diseases using leaf images.

He said advanced architectures such as ResNet50, DenseNet121, InceptionV3 and MobileNetV2 significantly improved detection performance but required substantial computational time.

The study found that techniques like hyperparameter tuning, transfer learning and data augmentation enhanced model accuracy while helping reduce training time.

It also revealed that larger datasets improve classification results, though they increase computational cost. The research emphasized the importance of balancing accuracy and efficiency to enable practical, large-scale adoption of automated plant disease diagnosis systems.

Co-supervisor of the scholar Dr Aftab Chandio, Dean Faculty of Natural Sciences Prof Dr Naheed Kaka, Dr Ayaz Keerio, Dr Saima Qayoom Memon, Dr Imtiaz Korejo, Dr Yasir Arfat, Dr Tariq Jalbani, Dr Asia Soomro, Dr Asadullah Buledi, Rafiq Mallah and others were also present on the occasion. External expert Dr Imtiaz Brohi attended the seminar via Zoom due to being present abroad.