Saifuddin Sagor, Md. Faysal Hossan, Hayder, Md Toha
This dataset comprises 1,625 original field images of seven economically important cotton insect pest species collected from multiple agricultural regions in Bangladesh, including Chandpur, Rajbari, Ashulia, Savar, and Dhaka, between 2024 and 2025. The dataset includes the following classes: Green Lacewings (300 images), Hadda Beetle (112 images), Hoverfly (108 images), Lady Beetle (438 images), Mirid Bug (305 images), Noctuidae (184 images), and Plant Bugs (178 images). All images were captured directly under natural field conditions using a vivo X27 smartphone camera (f/1.8 aperture, 1/328 s shutter speed). To improve the diversity and robustness of the dataset, images were acquired under varying environmental conditions, including different illumination levels, viewing angles, backgrounds, and pest orientations. The original images were standardized to a resolution of 560 × 420 pixels and stored in the sRGB color space, ensuring compatibility with a wide range of computer vision and deep learning frameworks. The dataset was developed to facilitate research on automated cotton insect pest detection, classification, and precision agriculture applications. It can be used for training, validation, benchmarking, and comparative evaluation of machine learning, deep learning, transfer learning, and vision transformer-based models. The diversity of acquisition conditions and multiple pest categories make the dataset suitable for developing robust and generalizable artificial intelligence models for real-world agricultural environments.