Abstract: In response to the high cost, slow speed, and difficulty in deploying deep learning models in weld defect detection, this paper proposes a lightweight and high precision optimized detection model (LHODM) with good performance. Firstly, build a new backbone network and design an ultra lightweight convolutional ULConv as the network foundation convolution to reduce model depth, generate more effective redundant feature maps using Depthwise operations, and reduce parameters and computational complexity. Secondly, design an efficient and lightweight dedicated module ELCC, which takes into account the distribution characteristics and imaging rules of weld defects under the premise of lightweight response model. Capture isolated area relationships from both horizontal and vertical directions, and combine with the lightweight upsampling operator CARAFE to make the model feature recombination have a larger receptive field, more effectively utilize environmental information, and compensate for the accuracy loss caused by lightweight. Finally, to improve convergence speed and loss function efficiency, an OS-IoU loss function is designed, taking into account the angle between the expected regression vectors, redefining the penalty term and correlation, and strengthening the attention to distance loss and shape loss. The results showed that the LHODM model achieved detection accuracy and speed of 91.62% and 63.47 frames per second, with a model memory cost of only 3.99 million, effectively solving the problems of high cost and slow speed in weld defect detection work.