<p>Underwater object detection is a prerequisite for underwater robots to achieve perception and operation. However, strong underwater background interference and weak underwater object perception greatly exacerbate the difficulty of underwater object detection. Furthermore, existing underwater image enhancement methods are mostly based on human vision and are difficult to effectively collaborate with computer vision tasks. Therefore, we propose a real-time learnable underwater adapter with color elimination and contrast enhancement, which aims to bridge the gap between underwater image enhancement and underwater object detection, and effectively reduce underwater background interference and improve underwater object perception. The color elimination module converts diverse color images into unified grayscale images to eliminate underwater color interference. The contrast enhancement module further processes the converted grayscale images into enhanced images to improve the distinction and recognition of underwater details. Both modules employ a self-learning process, which enhances target perception capability through point-wise and wavelet convolutions, and suppresses noise interference through spatial and channel attentions, in order to convert these images that conform to human visual perception into features that are easy for computer vision to understand. After autonomous learning, the two different branches focus on understanding underwater white and black target information respectively. These auxiliary learning features are stacked with the original image in the channel dimension and then passed into the YOLO detector together to complete feature encoding and feature decoding. Experimental results show that our method has extremely low cost and can achieve significant performance gains in underwater detection. Our code is available at <a href="https://github.com/shenxin-lnpc/LUA">https://github.com/shenxin-lnpc/LUA</a>.</p>