Marine environment monitoring by human being is not secure due to presence of different species. Mostly unmanned automated vehicles are used to monitor aquatic environment which needs a proper object detection model. Moreover underwater images are blur, so it is one of the challenging tasks to extract relevant features from images. This paper utilizes a deep learning object detection technique, based on YOLOv8, to create a network named ResCBAM-YOLOv8 for the purpose of detecting underwater objects. Residual Convolutional Block Attention Module (ResCBAM) consists of residual block with channel attention and spatial attention mechanism which helps to recalibrate feature maps within the residual blocks by adjusting their relevance at both the channel and spatial levels in an adaptive manner. The proposed work used Brackish dataset to perform underwater object detection and compared with YOLOv5, YOLOv7, YOLOv8. Precision, recall F1-score, and mAP used as assessment parameter, and it is found that proposed work achieved 2–7% better mAP than YOLOv5, YOLOv7, YOLOv8.